Вы находитесь на странице: 1из 482

Data Warehouse Concepts

Avinash Kanumuru Diya Jana Debyajit Majumder

2009 Wipro Ltd - Confidential

Content
1 An Overview of Data Warehouse 2 Data Warehouse Architecture 3 Data Modeling for Data Warehouse 4 Overview of Data Cleansing

5 Data Extraction, Transformation, Load

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Content [contd]
6 Metadata Management 7 OLAP 8 Data Warehouse Testing

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

An Overview
Understanding What is a Data Warehouse

2009 Wipro Ltd - Confidential

2009 Wipro Ltd - Confidential

What is Data Warehouse?


Definitions of Data Warehouse A data warehouse is a subject-oriented, integrated, nonvolatile, time-variant collection of data in support of management's decisions. WH Inmon Data Warehouse is a repository of data summarized or aggregated in simplified form from operational systems. End user orientated data access and reporting tools let user get at the data for decision support Babcock A data warehouse is a relational database a copy of transaction data specifically structured for query and analysis Ralph Kimball In simple: Data warehousing is collection of data from different systems, which helps in Business Decisions, Analysis and Reporting.

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Warehouse def. by WH Inmon


A common way of introducing data warehousing is to refer to the characteristics of a data warehouse as set forth by William Inmon: Subject Oriented Data that gives information about a particular subject instead of about a company's ongoing operations. Integrated Data that is gathered into the data warehouse from a variety of sources and merged into a coherent whole. Nonvolatile Data is stable in a data warehouse. More data is added but data is never removed. This enables management to gain a consistent picture of the business. Time Variant In order to discover trends in business, analysts need large amounts of data. This is very much in contrast to online transaction processing (OLTP) systems, where performance requirements demand that historical data be moved to an archive. All data in the data warehouse is identified with a particular time period.

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Warehouse Architecture


What makes a Data Warehouse

2009 Wipro Ltd - Confidential

Data Warehouse def. by WH Inmon


A common way of introducing data warehousing is to refer to the characteristics of a data warehouse as set forth by William Inmon: Subject Oriented Data that gives information about a particular subject instead of about a company's ongoing operations. Integrated Data that is gathered into the data warehouse from a variety of sources and merged into a coherent whole. Nonvolatile Data is stable in a data warehouse. More data is added but data is never removed. This enables management to gain a consistent picture of the business. Time Variant In order to discover trends in business, analysts need large amounts of data. This is very much in contrast to online transaction processing (OLTP) systems, where performance requirements demand that historical data be moved to an archive. All data in the data warehouse is identified with a particular time period.

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Components of Warehouse
Source Tables: These are real-time, volatile data in relational databases for transaction processing (OLTP). These can be any relational databases or flat files. ETL Tools: To extract, cleansing, transform (aggregates, joins) and load the data from sources to target. Maintenance and Administration Tools: To authorize and monitor access to the data, set-up users. Scheduling jobs to run on offshore periods. Modeling Tools: Used for data warehouse design for high-performance using dimensional data modeling technique, mapping the source and target files. Databases: Target databases and data marts, which are part of data warehouse. These are structured for analysis and reporting purposes. End-user tools for analysis and reporting: get the reports and analyze the data from target tables. Different types of Querying, Data Mining, OLAP tools are used for this purpose.

10

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Warehouse Architecture


This is a basic design, where there are source files, which are loaded to a warehouse and users query the data for different purposes.

This has a staging area, where the data after cleansing, transforming is loaded and tested here. Later is directly loaded to the target database/warehouse. Which is divided to data marts and can be accessed by different users for their reporting and analyzing purposes.

11

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Modeling
Effective way of using a Data Warehouse

12

2009 Wipro Ltd - Confidential

Data Warehouse def. by WH Inmon


A common way of introducing data warehousing is to refer to the characteristics of a data warehouse as set forth by William Inmon: Subject Oriented Data that gives information about a particular subject instead of about a company's ongoing operations. Integrated Data that is gathered into the data warehouse from a variety of sources and merged into a coherent whole. Nonvolatile Data is stable in a data warehouse. More data is added but data is never removed. This enables management to gain a consistent picture of the business. Time Variant In order to discover trends in business, analysts need large amounts of data. This is very much in contrast to online transaction processing (OLTP) systems, where performance requirements demand that historical data be moved to an archive. All data in the data warehouse is identified with a particular time period.

13

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Modeling
Commonly E-R Data Model is used in OLTP, In OLAP Dimensional Data Model is used commonly. E-R (Entity-Relationship) Data Model
Entity: Object that can be observed and classified based on its properties and characteristics. Like employee, book, student Relationship: relating entities to other entities.

Different Perceptive of Data Modeling.


o Conceptual Data Model o Logical Data Model o Physical Data Model

Types of Dimensional Data Models most commonly used:


o Star Schema o Snowflake Schema
14
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Terms used in Dimensional Data Model


To understand dimensional data modeling, let's define some of the terms commonly used in this type of modeling: Dimension: A category of information. For example, the time dimension. Attribute: A unique level within a dimension. For example, Month is an attribute in the Time Dimension. Hierarchy: The specification of levels that represents relationship between different attributes within a dimension. For example, one possible hierarchy in the Time dimension is Year Quarter Month Day. Fact Table: A table that contains the measures of interest. Lookup Table: It provides the detailed information about the attributes. For example, the lookup table for the Quarter attribute would include a list of all of the quarters available in the data warehouse. Surrogate Keys: To avoid the data integrity, surrogate keys are used. They are helpful for Slow Changing Dimensions and act as index/primary keys.
A dimensional model includes fact tables and lookup tables. Fact tables connect to one or more lookup tables, but fact tables do not have direct relationships to one another. Dimensions and hierarchies are represented by lookup tables. Attributes are the non-key columns in the lookup tables.
15
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Star Schema
Dimension Table
product prodId p1 p2 name price bolt 10 nut 5

Dimension Table
store storeId c1 c2 c3 city nyc sfo la

Fact Table
sale oderId date o100 1/7/97 o102 2/7/97 105 3/8/97 custId 53 53 111 prodId p1 p2 p1 storeId c1 c1 c3 qty 1 2 5 amt 12 11 50

Dimension Table
customer custId 53 81 111 name joe fred sally address 10 main 12 main 80 willow city sfo sfo la

16

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Warehouse def. by WH Inmon


A common way of introducing data warehousing is to refer to the characteristics of a data warehouse as set forth by William Inmon: Subject Oriented Data that gives information about a particular subject instead of about a company's ongoing operations. Integrated Data that is gathered into the data warehouse from a variety of sources and merged into a coherent whole. Nonvolatile Data is stable in a data warehouse. More data is added but data is never removed. This enables management to gain a consistent picture of the business. Time Variant In order to discover trends in business, analysts need large amounts of data. This is very much in contrast to online transaction processing (OLTP) systems, where performance requirements demand that historical data be moved to an archive. All data in the data warehouse is identified with a particular time period.

17

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Snowflake Schema
Dimension Table Fact Table
store storeId s5 s7 s9 cityId sfo sfo la tId t1 t2 t1 mgr joe fred nancy
sType tId t1 t2 city size small large location downtown suburbs regId north south

Dimension Table
cityId pop sfo 1M la 5M

The star and snowflake schema are most commonly found in dimensional data warehouses and data marts where speed of data retrieval is more important than the efficiency of data manipulations. As such, the tables in these schema are not normalized much, and are frequently designed at a level of normalization short of third normal form.

region regId name north cold region south warm region

18

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Overview of Data Cleansing

19

2009 Wipro Ltd - Confidential

The Need For Data Quality


Difficulty in decision making Time delays in operation Organizational mistrust Data ownership conflicts Customer attrition Costs associated with error detection error rework customer service fixing customer problems

20

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Warehouse def. by WH Inmon


A common way of introducing data warehousing is to refer to the characteristics of a data warehouse as set forth by William Inmon: Subject Oriented Data that gives information about a particular subject instead of about a company's ongoing operations. Integrated Data that is gathered into the data warehouse from a variety of sources and merged into a coherent whole. Nonvolatile Data is stable in a data warehouse. More data is added but data is never removed. This enables management to gain a consistent picture of the business. Time Variant In order to discover trends in business, analysts need large amounts of data. This is very much in contrast to online transaction processing (OLTP) systems, where performance requirements demand that historical data be moved to an archive. All data in the data warehouse is identified with a particular time period.

21

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Six Steps To Data Quality


Understand Information Flow In Organization
Identify authoritative data sources Interview Employees & Customers

Identify Potential Problem Areas & Asses Impact

Data Entry Points


Cost of bad data

Measure Quality Of Data

Use business rule discovery tools to identify data with

inconsistent, missing, incomplete, duplicate or incorrect values


Use data cleansing tools to clean data at the source Load only clean data into the data warehouse

Clean & Load Data

Continuous Monitoring

Schedule Periodic Cleansing of Source Data

Identify Areas of Improvement

Identify & Correct Cause of Defects Refine data capture mechanisms at source Educate users on importance of DQ
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

22

Data Quality Solution


Customized Programs Strengths: Addresses specific needs No bulky one time investment Limitations Tons of Custom programs in different environments are difficult to manage Minor alterations demand coding efforts Data Quality Assessment tools Strength Provide automated assessment Limitation No measure of data accuracy

23

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Quality Solution


Business Rule Discovery tools Strengths Detect Correlation in data values Can detect Patterns of behavior that indicate fraud Limitations Not all variables can be discovered Some discovered rules might not be pertinent There may be performance problems with large files or with many fields. Data Reengineering & Cleansing tools Strengths Usually are integrated packages with cleansing features as Addon Limitations Error prevention at source is usually absent The ETL tools have limited cleansing facilities
24
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Warehouse def. by WH Inmon


A common way of introducing data warehousing is to refer to the characteristics of a data warehouse as set forth by William Inmon: Subject Oriented Data that gives information about a particular subject instead of about a company's ongoing operations. Integrated Data that is gathered into the data warehouse from a variety of sources and merged into a coherent whole. Nonvolatile Data is stable in a data warehouse. More data is added but data is never removed. This enables management to gain a consistent picture of the business. Time Variant In order to discover trends in business, analysts need large amounts of data. This is very much in contrast to online transaction processing (OLTP) systems, where performance requirements demand that historical data be moved to an archive. All data in the data warehouse is identified with a particular time period.

25

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Tools In The Market


Business Rule Discovery Tools Integrity Data Reengineering Tool from Vality Technology Trillium Software System from Harte -Hanks Data Technologies Migration Architect from DB Star Data Reengineering & Cleansing Tools Carlton Pureview from Oracle ETI-Extract from Evolutionary Technologies PowerMart from Informatica Corp Sagent Data Mart from Sagent Technology Data Quality Assessment Tools Migration Architect, Evoke Axio from Evoke Software Wizrule from Wizsoft Name & Address Cleansing Tools Centrus Suite from Sagent I.d.centric from First Logic

26

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Extraction, Transformation, Load

27

2009 Wipro Ltd - Confidential

ETL Architecture

Visitors

Web Browsers

The Internet

External Data Demographics, Household, Webographics, Income

Staging Area
Web Server Logs & E-comm Transaction Data Flat Files

Meta Data Repository

Scheduled Extraction

RDBMS

Clean Transform Match Merge

Scheduled Loading

Enterprise Data Warehouse

Other OLTP Systems

Data Collection

Data Extraction

Data Transformation

Data Loading

Data Storage & Integration

28

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

ETL Architecture

Visitors

Web Browsers

The Internet

External Data Demographics, Household, Webographics, Income

Staging Area
Web Server Logs & E-comm Transaction Data Flat Files

Meta Data Repository

Scheduled Extraction

RDBMS

Clean Transform Match Merge

Scheduled Loading

Enterprise Data Warehouse

Other OLTP Systems

Data Collection

Data Extraction

Data Transformation

Data Loading

Data Storage & Integration

29

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

ETL Architecture

Visitors

Web Browsers

The Internet

External Data Demographics, Household, Webographics, Income

Staging Area
Web Server Logs & E-comm Transaction Data Flat Files

Meta Data Repository

Scheduled Extraction

RDBMS

Clean Transform Match Merge

Scheduled Loading

Enterprise Data Warehouse

Other OLTP Systems

Data Collection

Data Extraction

Data Transformation

Data Loading

Data Storage & Integration

30

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

ETL Architecture

Visitors

Web Browsers

The Internet

External Data Demographics, Household, Webographics, Income

Staging Area
Web Server Logs & E-comm Transaction Data Flat Files

Meta Data Repository

Scheduled Extraction

RDBMS

Clean Transform Match Merge

Scheduled Loading

Enterprise Data Warehouse

Other OLTP Systems

Data Collection

Data Extraction

Data Transformation

Data Loading

Data Storage & Integration

31

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

ETL Architecture

Visitors

Web Browsers

The Internet

External Data Demographics, Household, Webographics, Income

Staging Area
Web Server Logs & E-comm Transaction Data Flat Files

Meta Data Repository

Scheduled Extraction

RDBMS

Clean Transform Match Merge

Scheduled Loading

Enterprise Data Warehouse

Other OLTP Systems

Data Collection

Data Extraction

Data Transformation

Data Loading

Data Storage & Integration

32

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

ETL Architecture

Visitors

Web Browsers

The Internet

External Data Demographics, Household, Webographics, Income

Staging Area
Web Server Logs & E-comm Transaction Data Flat Files

Meta Data Repository

Scheduled Extraction

RDBMS

Clean Transform Match Merge

Scheduled Loading

Enterprise Data Warehouse

Other OLTP Systems

Data Collection

Data Extraction

Data Transformation

Data Loading

Data Storage & Integration

33

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

ETL Architecture

Visitors

Web Browsers

The Internet

External Data Demographics, Household, Webographics, Income

Staging Area
Web Server Logs & E-comm Transaction Data Flat Files

Meta Data Repository

Scheduled Extraction

RDBMS

Clean Transform Match Merge

Scheduled Loading

Enterprise Data Warehouse

Other OLTP Systems

Data Collection

Data Extraction

Data Transformation

Data Loading

Data Storage & Integration

34

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

ETL Architecture

Visitors

Web Browsers

The Internet

External Data Demographics, Household, Webographics, Income

Staging Area
Web Server Logs & E-comm Transaction Data Flat Files

Meta Data Repository

Scheduled Extraction

RDBMS

Clean Transform Match Merge

Scheduled Loading

Enterprise Data Warehouse

Other OLTP Systems

Data Collection

Data Extraction

Data Transformation

Data Loading

Data Storage & Integration

35

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

ETL Architecture
Data Extraction:
Rummages through a file or database Uses some criteria for selection Identifies qualified data and Transports the data over onto another file or database

Data transformation
Integrating dissimilar data types Changing codes Adding a time attribute Summarizing data Calculating derived values Renormalizing data

Data Extraction Cleanup


Restructuring of records or fields Removal of Operational-only data Supply of missing field values Data Integrity checks Data Consistency and Range checks, etc...

Data loading
Initial and incremental loading Updation of metadata

36

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Why ETL ?
Companies have valuable data lying around throughout their networks that needs to be moved from one place to another. The data lies in all sorts of heterogeneous systems,and therefore in all sorts of formats.

To solve the problem, companies use extract, transform and load (ETL) software.
The data used in ETL processes can come from any source: a mainframe application, an ERP application, a CRM tool, a flat file, and an Excel spreadsheet.

37

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

ETL Architecture

Visitors

Web Browsers

The Internet

External Data Demographics, Household, Webographics, Income

Staging Area
Web Server Logs & E-comm Transaction Data Flat Files

Meta Data Repository

Scheduled Extraction

RDBMS

Clean Transform Match Merge

Scheduled Loading

Enterprise Data Warehouse

Other OLTP Systems

Data Collection

Data Extraction

Data Transformation

Data Loading

Data Storage & Integration

38

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Major components involved in ETL Processing

39

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Major components involved in ETL Processing


Design manager Lets developers define source-to-target mappings, transformations, process flows, and jobs Meta data management Provides a repository to define, document, and manage information about the ETL design and runtime processes Extract The process of reading data from a database. Transform The process of converting the extracted data Load The process of writing the data into the target database. Transport services ETL tools use network and file protocols to move data between source and target systems and in-memory protocols to move data between ETL run-time components. Administration and operation ETL utilities let administrators schedule, run, monitor ETL jobs, log all events, manage errors, recover from failures, reconcile outputs with source systems
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

40

ETL Tools
Provides facility to specify a large number of transformation rules with a GUI Generate programs to transform data Handle multiple data sources Handle data redundancy Generate metadata as output Most tools exploit parallelism by running on multiple low-cost servers in multi-threaded environment ETL Tools - Second-Generation PowerCentre/Mart from Informatica Data Mart Solution from Sagent Technology DataStage from Ascential
41

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

ETL Architecture

Visitors

Web Browsers

The Internet

External Data Demographics, Household, Webographics, Income

Staging Area
Web Server Logs & E-comm Transaction Data Flat Files

Meta Data Repository

Scheduled Extraction

RDBMS

Clean Transform Match Merge

Scheduled Loading

Enterprise Data Warehouse

Other OLTP Systems

Data Collection

Data Extraction

Data Transformation

Data Loading

Data Storage & Integration

42

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Metadata Management

43

2009 Wipro Ltd - Confidential

What Is Metadata?
Metadata is Information...

That describes the WHAT, WHEN, WHO, WHERE, HOW of the data warehouse About the data being captured and loaded into the Warehouse Documented in IT tools that improves both business and technical understanding of data and data-related processes

44

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Importance Of Metadata
Locating Information Time spent in looking for information. How often information is found? What poor decisions were made based on the incomplete information?

How much money was lost or earned as a result? Interpreting information


How many times have businesses needed to rework or recall products? What impact does it have on the bottom line ? How many mistakes were due to misinterpretation of existing How much interpretation results form too much metadata? How much time is spent trying to determine if any of the metadata is accurate? Integrating information How various data perspectives connect together? How much time is spent trying to figure out that? How much does the inefficiency and lack of metadata affect decision making documentation?

45

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

ETL Architecture

Visitors

Web Browsers

The Internet

External Data Demographics, Household, Webographics, Income

Staging Area
Web Server Logs & E-comm Transaction Data Flat Files

Meta Data Repository

Scheduled Extraction

RDBMS

Clean Transform Match Merge

Scheduled Loading

Enterprise Data Warehouse

Other OLTP Systems

Data Collection

Data Extraction

Data Transformation

Data Loading

Data Storage & Integration

46

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Requirements for DW Metadata Management


Provide a simple catalogue of business metadata descriptions and views Document/manage metadata descriptions from an integrated development environment Enable DW users to identify and invoke pre-built queries against the data stores Design and enhance new data models and schemas for the data warehouse Capture data transformation rules between the operational and data warehousing databases Provide change impact analysis, and update across these technologies
47
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Consumers of Metadata
Technical Users Warehouse administrator Application developer Business Users -Business metadata Meanings Definitions Business Rules Software Tools Used in DW life-cycle development Metadata requirements for each tool must be identified The tool-specific metadata should be analysed for inclusion in the enterprise metadata repository Previously captured metadata should be electronically transferred from the enterprise metadata repository to each individual tool

48

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Trends in the Metadata Management Tools


Third Party Bridging Tools Oracle Exchange
Technology of choice for a long list of repository, enterprise and workgroup vendors

Reischmann-Informatik-Toolbus
Features include facilitation of selective bridging of metadata

Ardent Software/ Dovetail Software -Interplay


Hub and Spoke solution for enabling metadata interoperability Ardent focussing on own engagements, not selling it as independent product

Informix's Metadata Plug-ins


Available with Ardent Datastage version 3.6.2 free of cost for Erwin, Oracle Designer, Sybase Powerdesigner, Brio, Microstrategy
49
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Trends in the Metadata Management Tools


Metadata Repositories IBM, Oracle and Microsoft to offer free or near-free basic repository services Enable organisations to reuse metadata across technologies Integrate DB design, data transformation and BI tools from different vendors Multi-tool vendors taking a bridged or federated rather than integrated approach to sharing metadata Both IBM and Oracle have multiple repositories for different lines of products e.g., One for AD and one for DW, with bridges between them

50

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Trends in the Metadata Management Tools


Metadata Interchange Standards CDIF (CASE Data Interchange Format)
Most frequently used interchange standard Addresses only a limited subset of metadata artifacts

OMG (Object Management Group)-CWM


XML-addresses context and data meaning, not presentation Can enable exchange over the web employing industry standards for storing and sharing programming data Will allow sharing of UML and MOF objects b/w various development tools and repositories

MDC (Metadata Coalition)


Based on XML/UML standards Promoted by Microsoft Along With 20 partners including Object Management Group (OMG), Oracle Carleton Group, CA-PLATINUM Technology (Founding Member), Viasoft
51
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

OLAP

52

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Agenda
OLAP Definition Distinction between OLTP and OLAP

MDDB Concepts
Implementation Techniques Architectures

Features
Representative Tools

6/19/2012

53

53

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

OLAP: On-Line Analytical Processing


OLAP can be defined as a technology which allows the users to view the aggregate data across measurements (like Maturity Amount, Interest Rate etc.) along with a set of related parameters called dimensions (like Product, Organization, Customer, etc.) Used interchangeably with BI Multidimensional view of data is the foundation of OLAP Users :Analysts, Decision makers
6/19/2012 54

54

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Distinction between OLTP and OLAP


OLTP System Source of data Operational data; OLTPs are the original source of the data To control and run fundamental business tasks A snapshot of ongoing business processes Short and fast inserts and updates initiated by end users
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

OLAP System Consolidation data; OLAP data comes from the various OLTP databases Decision support

Purpose of data

What the data reveals Inserts and Updates


6/19/2012

Multi-dimensional views of various kinds of business activities Periodic long-running batch jobs refresh the data
55

55

MDDB Concepts
A multidimensional database is a computer software system designed to allow for efficient and convenient storage and retrieval of data that is intimately related and stored, viewed and analyzed from different perspectives (Dimensions). A hypercube represents a collection of multidimensional data. The edges of the cube are called dimensions Individual items within each dimensions are called members
56
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

RDBMS v/s MDDB: Increased Complexity...


Relational DBMS
MODEL MINI VAN MINI VAN MINI VAN MINI VAN MINI VAN MINI VAN MINI VAN MINI VAN MINI VAN SPORTS COUPE SPORTS COUPE SPORTS COUPE SPORTS COUPE SPORTS COUPE SPORTS COUPE SPORTS COUPE SPORTS COUPE SPORTS COUPE SEDAN SEDAN SEDAN ... COLOR BLUE BLUE BLUE RED RED RED WHITE WHITE WHITE BLUE BLUE BLUE RED RED RED WHITE WHITE WHITE BLUE BLUE BLUE DEALER Clyde Gleason Carr Clyde Gleason Carr Clyde Gleason Carr Clyde Gleason Carr Clyde Gleason Carr Clyde Gleason Carr Clyde Gleason Carr VOL. 6 3 2 5 3 1 3 1 4 3 3 3 4 3 6 2 3 5 4 3 2 ...

MDDB

Sales Volumes

M O D E L

Mini Van

Coupe Carr Gleason Clyde Blue Red White

Sedan

DEALERSHIP

COLOR

27 x 4 = 108 cells
57

3 x 3 x 3 = 27 cells

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Benefits of MDDB over RDBMS


Ease of Data Presentation & Navigation A great deal of information is gleaned immediately upon direct inspection of the array User is able to view data along presorted dimensions with data arranged in an inherently more organized, and accessible fashion than the one offered by the relational table. Storage Space Very low Space Consumption compared to Relational DB Performance Gives much better performance. Relational DB may give comparable results only through database tuning (indexing, keys etc), which may not be possible for ad-hoc queries. Ease of Maintenance No overhead as data is stored in the same way it is viewed. In Relational DB, indexes, sophisticated joins etc. are used which require considerable storage and maintenance
6/19/2012
58
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

58

Issues with MDDB

Sparsity
- Input data in applications are typically sparse -Increases with increased dimensions

Data Explosion
-Due to Sparsity -Due to Summarization

Performance
-Doesnt perform better than RDBMS at high data volumes (>20-30 GB)

6/19/2012
59
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

59

Issues with MDDB - Sparsity Example


If dimension members of different dimensions do not interact , then blank cell is left behind.
Employee Age
21 19 63 31 27 56 45 41 19
31 41 23 01 14 54 03 12 33

LAST NAME EMP# AGE SMITH 01 21 REGAN 12 Sales Volumes 19 FOX 31 63 Miini Van WELD 14 6 5 31 4 M O KELLY 54 3 5 27 Coupe D 5 E L LINK 03 56 4 3 2 Sedan KRANZ 41 45 Blue Red White LUCUS 33 COLOR 41 WEISS 23 19

Smith

Regan

Fox

L A S T N A M E

Weld

Kelly

Link

Kranz

Lucas

Weiss

EMPLOYEE #

6/19/2012
60
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

60

OLAP Features
Calculations applied across dimensions, through hierarchies and/or across members Trend analysis over sequential time periods, What-if scenarios. Slicing / Dicing subsets for on-screen viewing Rotation to new dimensional comparisons in the viewing area Drill-down/up along the hierarchy Reach-through / Drill-through to underlying detail data

6/19/2012
61
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

61

Features of OLAP - Rotation

Complex Queries & Sorts in Relational environment translated to simple rotation.


Sales Volumes

M O D E L

Mini Van

6 3 4
Blue

5 5 3
Red

4 5 2
White

Coupe

C O L O R ( ROTATE 90 )
o

Blue

6 5 4

3 5 5
MODEL

4 3 2
Sedan

Red

Sedan

White

Mini Van Coupe

COLOR

View #1

View #2

2 dimensional array has 2 views.


6/19/2012
62
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

62

Features of OLAP - Rotation


Sales Volumes

M O D E L

Mini Van Coupe Carr Gleason Clyde Blue Red White

Sedan

C O L O R

Blue

Red White Sedan Coupe Mini Van Carr Gleason Clyde

C O L O R

Blue

Red White Carr Gleason Clyde Mini Van Coupe Sedan

COLOR

( ROTATE 90 )

MODEL

( ROTATE 90 )

DEALERSHIP

( ROTATE 90 )

DEALERSHIP

DEALERSHIP

MODEL

View #1
D E A L E R S H I P D E A L E R S H I P

View #2

View #3

Carr Gleason Mini Van Coupe Sedan White Red Blue

Carr Gleason Blue Red White

Mini Van

Clyde

Clyde Mini Van Coupe Sedan

M O D E L

Coupe Blue Red White Clyde Gleason Carr

Sedan

COLOR

( ROTATE 90 )

MODEL

( ROTATE 90 )

DEALERSHIP

MODEL

COLOR

COLOR

View #4

View #5

View #6

3 dimensional array has 6 views.


6/19/2012
63
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

63

Features of OLAP - Slicing / Filtering


MDDB allows end user to quickly slice in on exact view of the data required.

Sales Volumes

M O D E L

Mini Van Mini Van

Coupe

Coupe Normal Metal Blue Blue

Carr Clyde

Carr Clyde

Normal Blue

Metal Blue

DEALERSHIP

COLOR
6/19/2012
64
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

64

Features of OLAP - Drill Down / Up

ORGANIZATION DIMENSION
REGION Midwest

DISTRICT

Chicago

St. Louis

Gary

DEALERSHIP

Clyde

Gleason

Carr

Levi

Lucas

Bolton

Sales at region/District/Dealership Level

Moving Up and moving down in a hierarchy is referred to as drill-up / roll-up and drill-down

6/19/2012
65
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

65

OLAP Reporting - Drill Down

Inflows ( Region , Year)


200 150 Inflows 100 ($M) 50 0 Year Year 1999 2000 Years

East West Central

6/19/2012
66
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

66

OLAP Reporting - Drill Down

Inflows ( Region , Year - Year 1999)


90 80 70 60 50 Inflows ( $M) 40 30 20 10 0

East West Central

1st Qtr

2nd Qtr 3rd Qtr Year 1999

4th Qtr

Drill-down from Year to Quarter


6/19/2012
67
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

67

OLAP Reporting - Drill Down

Inflows ( Region , Year - Year 1999 - 1st Qtr)


20 15 Inflows ( $M 10 ) 5

East West Central


January February March Year 1999

Drill-down from Quarter to Month

68

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Implementation Techniques -OLAP Architectures

MOLAP - Multidimensional OLAP


Multidimensional Databases for database and application logic layer

ROLAP - Relational OLAP


Access Data stored in relational Data Warehouse for OLAP Analysis. Database and Application logic provided as separate layers

HOLAP - Hybrid OLAP


OLAP Server routes queries first to MDDB, then to RDBMS and result processed on-the-fly in Server

DOLAP - Desk OLAP


Personal MDDB Server and application on the desktop

6/19/2012
69
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

69

MOLAP - MDDB storage

OLAP
Cube
OLAP Calculation Engine

Web Browser

OLAP Tools

OLAP Appli cations


6/19/2012
70
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

70

MOLAP - Features

Powerful analytical capabilities (e.g., financial, forecasting, statistical) Aggregation and calculation capabilities Read/write analytic applications Specialized data structures for
Maximum query performance. Optimum space utilization.
6/19/2012
71
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

71

ROLAP - Standard SQL storage

MDDB - Relational Mapping

Relational DW

Web Browser
OLAP Calculation Engine

SQL

OLAP Tools

OLAP Applications
6/19/2012
72
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

72

ROLAP - Features Three-tier hardware/software architecture:


GUI on client; multidimensional processing on midtier server; target database on database server Processing split between mid-tier & database servers

Ad hoc query capabilities to very large databases DW integration Data scalability

6/19/2012
73
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

73

HOLAP - Combination of RDBMS and MDDB


OLAP Cube
Any Client

Relational DW

Web Browser
OLAP Calculation Engine

SQL

OLAP Tools

OLAP Applications
6/19/2012
74
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

74

HOLAP - Features

RDBMS used for detailed data stored in large databases MDDB used for fast, read/write OLAP analysis and calculations Scalability of RDBMS and MDDB performance Calculation engine provides full analysis features Source of data transparent to end user

6/19/2012
75
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

75

Architecture Comparison

MOLAP
Definition

ROLAP

HOLAP
Hybrid OLAP = ROLAP + summary in MDDB Sparsity exists only in MDDB part To the necessary extent

MDDB OLAP = Relational OLAP = Transaction level data + Transaction level data + summary in MDDB summary in RDBMS Good Design 3 10 times High (May go beyond control. Estimation is very important) Fast - (Depends upon the size of the MDDB) No Sparsity To the necessary extent

Data explosion due to Sparsity Data explosion due to Summarization Query Execution Speed

Slow

Optimum - If the data is fetched from RDBMS then its like ROLAP otherwise like MOLAP. High: RDBMS + disk space + MDDB Server cost Large transactional data + frequent summary analysis

Cost

Medium: MDDB Server + large disk space cost

Low: Only RDBMS + disk space cost

Where to apply?

Small transactional Very large transactional data + complex model + data & it needs to be frequent summary viewed / sorted analysis

6/19/2012
76
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

76

Representative OLAP Tools:

Oracle Express Products Hyperion Essbase Cognos -PowerPlay Seagate - Holos SAS

Micro Strategy - DSS Agent Informix MetaCube Brio Query Business Objects / Web Intelligence

6/19/2012
77
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

77

Sample OLAP Applications

Sales Analysis Financial Analysis Profitability Analysis Performance Analysis Risk Management Profiling & Segmentation Scorecard Application NPA Management Strategic Planning Customer Relationship Management (CRM)
6/19/2012
78
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

78

Data Warehouse Testing

79

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Warehouse Testing Overview


There is an exponentially increasing cost associated with finding software defects later in the development lifecycle. In data warehousing, this is compounded because of the additional business costs of using incorrect data to make critical business decisions

The methodology required for testing a Data Warehouse is different from testing a typical transaction system

80

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Difference In Testing Data warehouse and Transaction System


Data warehouse testing is different on the following counts: User-Triggered vs. System triggered Volume of Test Data Possible scenarios/ Test Cases Programming for testing challenge

81

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Difference In Testing Data warehouse and Transaction System.


User-Triggered vs. System triggered

In data Warehouse, most of the testing is system triggered. Most of the production/Source system testing is the processing of individual transactions, which are driven by some input from the users (Application Form, Servicing Request.). There are very few test cycles, which cover the system-triggered scenarios (Like billing, Valuation.)

82

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Difference In Testing Data warehouse and Transaction System


Volume of Test Data The test data in a transaction system is a very small sample of the overall production data. Data Warehouse has typically large test data as one does try to fill-up maximum possible combination of dimensions and facts. Possible scenarios/ Test Cases In case of Data Warehouse, the permutations and combinations one can possibly test is virtually unlimited due to the core objective of Data Warehouse is to allow all possible views of data.

83

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Difference In Testing Data warehouse and Transaction System


Programming for testing challenge In case of transaction systems, users/business analysts typically test the output of the system. In case of data warehouse, most of the 'Data Warehouse data Quality testing' and ETL testing is done at backend by running separate stand-alone scripts. These scripts compare preTransformation to post Transformation of data.

84

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Warehouse Testing Process


Data-Warehouse testing is basically divided into two parts : 'Back-end' testing where the source systems data is compared to the endresult data in Loaded area 'Front-end' testing where the user checks the data by comparing their MIS with the data displayed by the end-user tools like OLAP. Testing phases consists of : Requirements testing Unit testing Integration testing Performance testing Acceptance testing

85

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Requirements testing
The main aim for doing Requirements testing is to check stated requirements for completeness. Requirements can be tested on following factors. Are the requirements Complete? Are the requirements Singular? Are the requirements Ambiguous? Are the requirements Developable? Are the requirements Testable?

86

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Unit Testing
Unit testing for data warehouses is WHITEBOX. It should check the ETL procedures/mappings/jobs and the reports developed. Unit testing the ETL procedures: Whether ETLs are accessing and picking up right data from right source.

All the data transformations are correct according to the business rules and data warehouse is correctly populated with the transformed data.
Testing the rejected records that dont fulfil transformation rules.

87

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Unit Testing
Unit Testing the Report data:

Verify Report data with source: Data present in a data warehouse will be stored at an aggregate level compare to source systems. QA team should verify the granular data stored in data warehouse against the source data available Field level data verification: QA team must understand the linkages for the fields displayed in the report and should trace back and compare that with the source systems Derivation formulae/calculation rules should be verified

88

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Integration Testing
Integration testing will involve following:

Sequence of ETLs jobs in batch. Initial loading of records on data warehouse. Incremental loading of records at a later date to verify the newly inserted or updated data. Testing the rejected records that dont fulfil transformation rules. Error log generation

89

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Performance Testing
Performance Testing should check for : ETL processes completing within time window.

Monitoring and measuring the data quality issues.


Refresh times for standard/complex reports.

90

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Acceptance testing
Here the system is tested with full functionality and is expected to function as in production. At the end of UAT, the system should be acceptable to the client for use in terms of ETL process integrity and business functionality and reporting.

91

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Questions

92

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Thank You

93

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Warehouse Concepts

Avinash Kanumuru Diya Jana Debyajit Majumder

2009 Wipro Ltd - Confidential

Content
1 An Overview of Data Warehouse 2 Data Warehouse Architecture 3 Data Modeling for Data Warehouse 4 Overview of Data Cleansing

5 Data Extraction, Transformation, Load

95

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Content [contd]
6 Metadata Management 7 OLAP 8 Data Warehouse Testing

96

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

An Overview
Understanding What is a Data Warehouse

97

2009 Wipro Ltd - Confidential

What is Data Warehouse?


Definitions of Data Warehouse A data warehouse is a subject-oriented, integrated, nonvolatile, time-variant collection of data in support of management's decisions. WH Inmon Data Warehouse is a repository of data summarized or aggregated in simplified form from operational systems. End user orientated data access and reporting tools let user get at the data for decision support Babcock A data warehouse is a relational database a copy of transaction data specifically structured for query and analysis Ralph Kimball In simple: Data warehousing is collection of data from different systems, which helps in Business Decisions, Analysis and Reporting.

98

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Warehouse def. by WH Inmon


A common way of introducing data warehousing is to refer to the characteristics of a data warehouse as set forth by William Inmon: Subject Oriented Data that gives information about a particular subject instead of about a company's ongoing operations. Integrated Data that is gathered into the data warehouse from a variety of sources and merged into a coherent whole. Nonvolatile Data is stable in a data warehouse. More data is added but data is never removed. This enables management to gain a consistent picture of the business. Time Variant In order to discover trends in business, analysts need large amounts of data. This is very much in contrast to online transaction processing (OLTP) systems, where performance requirements demand that historical data be moved to an archive. All data in the data warehouse is identified with a particular time period.

99

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Warehouse Architecture


What makes a Data Warehouse

100

2009 Wipro Ltd - Confidential

Data Warehouse Concepts

Avinash Kanumuru Diya Jana Debyajit Majumder

2009 Wipro Ltd - Confidential

Content
1 An Overview of Data Warehouse 2 Data Warehouse Architecture 3 Data Modeling for Data Warehouse 4 Overview of Data Cleansing

5 Data Extraction, Transformation, Load

102

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Content [contd]
6 Metadata Management 7 OLAP 8 Data Warehouse Testing

103

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

An Overview
Understanding What is a Data Warehouse

104

2009 Wipro Ltd - Confidential

What is Data Warehouse?


Definitions of Data Warehouse A data warehouse is a subject-oriented, integrated, nonvolatile, time-variant collection of data in support of management's decisions. WH Inmon Data Warehouse is a repository of data summarized or aggregated in simplified form from operational systems. End user orientated data access and reporting tools let user get at the data for decision support Babcock A data warehouse is a relational database a copy of transaction data specifically structured for query and analysis Ralph Kimball In simple: Data warehousing is collection of data from different systems, which helps in Business Decisions, Analysis and Reporting.

105

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Warehouse def. by WH Inmon


A common way of introducing data warehousing is to refer to the characteristics of a data warehouse as set forth by William Inmon: Subject Oriented Data that gives information about a particular subject instead of about a company's ongoing operations. Integrated Data that is gathered into the data warehouse from a variety of sources and merged into a coherent whole. Nonvolatile Data is stable in a data warehouse. More data is added but data is never removed. This enables management to gain a consistent picture of the business. Time Variant In order to discover trends in business, analysts need large amounts of data. This is very much in contrast to online transaction processing (OLTP) systems, where performance requirements demand that historical data be moved to an archive. All data in the data warehouse is identified with a particular time period.

106

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Warehouse Architecture


What makes a Data Warehouse

107

2009 Wipro Ltd - Confidential

Data Warehouse Concepts

Avinash Kanumuru Diya Jana Debyajit Majumder

2009 Wipro Ltd - Confidential

Content
1 An Overview of Data Warehouse 2 Data Warehouse Architecture 3 Data Modeling for Data Warehouse 4 Overview of Data Cleansing

5 Data Extraction, Transformation, Load

109

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Content [contd]
6 Metadata Management 7 OLAP 8 Data Warehouse Testing

110

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

An Overview
Understanding What is a Data Warehouse

111

2009 Wipro Ltd - Confidential

What is Data Warehouse?


Definitions of Data Warehouse A data warehouse is a subject-oriented, integrated, nonvolatile, time-variant collection of data in support of management's decisions. WH Inmon Data Warehouse is a repository of data summarized or aggregated in simplified form from operational systems. End user orientated data access and reporting tools let user get at the data for decision support Babcock A data warehouse is a relational database a copy of transaction data specifically structured for query and analysis Ralph Kimball In simple: Data warehousing is collection of data from different systems, which helps in Business Decisions, Analysis and Reporting.

112

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Warehouse def. by WH Inmon


A common way of introducing data warehousing is to refer to the characteristics of a data warehouse as set forth by William Inmon: Subject Oriented Data that gives information about a particular subject instead of about a company's ongoing operations. Integrated Data that is gathered into the data warehouse from a variety of sources and merged into a coherent whole. Nonvolatile Data is stable in a data warehouse. More data is added but data is never removed. This enables management to gain a consistent picture of the business. Time Variant In order to discover trends in business, analysts need large amounts of data. This is very much in contrast to online transaction processing (OLTP) systems, where performance requirements demand that historical data be moved to an archive. All data in the data warehouse is identified with a particular time period.

113

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Warehouse Architecture


What makes a Data Warehouse

114

2009 Wipro Ltd - Confidential

Components of Warehouse
Source Tables: These are real-time, volatile data in relational databases for transaction processing (OLTP). These can be any relational databases or flat files. ETL Tools: To extract, cleansing, transform (aggregates, joins) and load the data from sources to target. Maintenance and Administration Tools: To authorize and monitor access to the data, set-up users. Scheduling jobs to run on offshore periods. Modeling Tools: Used for data warehouse design for high-performance using dimensional data modeling technique, mapping the source and target files. Databases: Target databases and data marts, which are part of data warehouse. These are structured for analysis and reporting purposes. End-user tools for analysis and reporting: get the reports and analyze the data from target tables. Different types of Querying, Data Mining, OLAP tools are used for this purpose.

115

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Warehouse Architecture


This is a basic design, where there are source files, which are loaded to a warehouse and users query the data for different purposes.

This has a staging area, where the data after cleansing, transforming is loaded and tested here. Later is directly loaded to the target database/warehouse. Which is divided to data marts and can be accessed by different users for their reporting and analyzing purposes.

116

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Modeling
Effective way of using a Data Warehouse

117

2009 Wipro Ltd - Confidential

Data Modeling
Commonly E-R Data Model is used in OLTP, In OLAP Dimensional Data Model is used commonly. E-R (Entity-Relationship) Data Model
Entity: Object that can be observed and classified based on its properties and characteristics. Like employee, book, student Relationship: relating entities to other entities.

Different Perceptive of Data Modeling.


o Conceptual Data Model o Logical Data Model o Physical Data Model

Types of Dimensional Data Models most commonly used:


o Star Schema o Snowflake Schema
118
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Terms used in Dimensional Data Model


To understand dimensional data modeling, let's define some of the terms commonly used in this type of modeling: Dimension: A category of information. For example, the time dimension. Attribute: A unique level within a dimension. For example, Month is an attribute in the Time Dimension. Hierarchy: The specification of levels that represents relationship between different attributes within a dimension. For example, one possible hierarchy in the Time dimension is Year Quarter Month Day. Fact Table: A table that contains the measures of interest. Lookup Table: It provides the detailed information about the attributes. For example, the lookup table for the Quarter attribute would include a list of all of the quarters available in the data warehouse. Surrogate Keys: To avoid the data integrity, surrogate keys are used. They are helpful for Slow Changing Dimensions and act as index/primary keys.
A dimensional model includes fact tables and lookup tables. Fact tables connect to one or more lookup tables, but fact tables do not have direct relationships to one another. Dimensions and hierarchies are represented by lookup tables. Attributes are the non-key columns in the lookup tables.
119
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Star Schema
Dimension Table
product prodId p1 p2 name price bolt 10 nut 5

Dimension Table
store storeId c1 c2 c3 city nyc sfo la

Fact Table
sale oderId date o100 1/7/97 o102 2/7/97 105 3/8/97 custId 53 53 111 prodId p1 p2 p1 storeId c1 c1 c3 qty 1 2 5 amt 12 11 50

Dimension Table
customer custId 53 81 111 name joe fred sally address 10 main 12 main 80 willow city sfo sfo la

120

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Snowflake Schema
Dimension Table Fact Table
store storeId s5 s7 s9 cityId sfo sfo la tId t1 t2 t1 mgr joe fred nancy
sType tId t1 t2 city size small large location downtown suburbs regId north south

Dimension Table
cityId pop sfo 1M la 5M

The star and snowflake schema are most commonly found in dimensional data warehouses and data marts where speed of data retrieval is more important than the efficiency of data manipulations. As such, the tables in these schema are not normalized much, and are frequently designed at a level of normalization short of third normal form.

region regId name north cold region south warm region

121

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Overview of Data Cleansing

122

2009 Wipro Ltd - Confidential

The Need For Data Quality


Difficulty in decision making Time delays in operation Organizational mistrust Data ownership conflicts Customer attrition Costs associated with error detection error rework customer service fixing customer problems

123

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Six Steps To Data Quality


Understand Information Flow In Organization
Identify authoritative data sources Interview Employees & Customers

Identify Potential Problem Areas & Asses Impact

Data Entry Points


Cost of bad data

Measure Quality Of Data

Use business rule discovery tools to identify data with

inconsistent, missing, incomplete, duplicate or incorrect values


Use data cleansing tools to clean data at the source Load only clean data into the data warehouse

Clean & Load Data

Continuous Monitoring

Schedule Periodic Cleansing of Source Data

Identify Areas of Improvement

Identify & Correct Cause of Defects Refine data capture mechanisms at source Educate users on importance of DQ
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

124

Data Quality Solution


Customized Programs Strengths: Addresses specific needs No bulky one time investment Limitations Tons of Custom programs in different environments are difficult to manage Minor alterations demand coding efforts Data Quality Assessment tools Strength Provide automated assessment Limitation No measure of data accuracy

125

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Quality Solution


Business Rule Discovery tools Strengths Detect Correlation in data values Can detect Patterns of behavior that indicate fraud Limitations Not all variables can be discovered Some discovered rules might not be pertinent There may be performance problems with large files or with many fields. Data Reengineering & Cleansing tools Strengths Usually are integrated packages with cleansing features as Addon Limitations Error prevention at source is usually absent The ETL tools have limited cleansing facilities
126
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Tools In The Market


Business Rule Discovery Tools Integrity Data Reengineering Tool from Vality Technology Trillium Software System from Harte -Hanks Data Technologies Migration Architect from DB Star Data Reengineering & Cleansing Tools Carlton Pureview from Oracle ETI-Extract from Evolutionary Technologies PowerMart from Informatica Corp Sagent Data Mart from Sagent Technology Data Quality Assessment Tools Migration Architect, Evoke Axio from Evoke Software Wizrule from Wizsoft Name & Address Cleansing Tools Centrus Suite from Sagent I.d.centric from First Logic

127

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Extraction, Transformation, Load

128

2009 Wipro Ltd - Confidential

ETL Architecture

Visitors

Web Browsers

The Internet

External Data Demographics, Household, Webographics, Income

Staging Area
Web Server Logs & E-comm Transaction Data Flat Files

Meta Data Repository

Scheduled Extraction

RDBMS

Clean Transform Match Merge

Scheduled Loading

Enterprise Data Warehouse

Other OLTP Systems

Data Collection

Data Extraction

Data Transformation

Data Loading

Data Storage & Integration

129

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

ETL Architecture
Data Extraction:
Rummages through a file or database Uses some criteria for selection Identifies qualified data and Transports the data over onto another file or database

Data transformation
Integrating dissimilar data types Changing codes Adding a time attribute Summarizing data Calculating derived values Renormalizing data

Data Extraction Cleanup


Restructuring of records or fields Removal of Operational-only data Supply of missing field values Data Integrity checks Data Consistency and Range checks, etc...

Data loading
Initial and incremental loading Updation of metadata

130

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Why ETL ?
Companies have valuable data lying around throughout their networks that needs to be moved from one place to another. The data lies in all sorts of heterogeneous systems,and therefore in all sorts of formats.

To solve the problem, companies use extract, transform and load (ETL) software.
The data used in ETL processes can come from any source: a mainframe application, an ERP application, a CRM tool, a flat file, and an Excel spreadsheet.

131

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Major components involved in ETL Processing

132

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Major components involved in ETL Processing


Design manager Lets developers define source-to-target mappings, transformations, process flows, and jobs Meta data management Provides a repository to define, document, and manage information about the ETL design and runtime processes Extract The process of reading data from a database. Transform The process of converting the extracted data Load The process of writing the data into the target database. Transport services ETL tools use network and file protocols to move data between source and target systems and in-memory protocols to move data between ETL run-time components. Administration and operation ETL utilities let administrators schedule, run, monitor ETL jobs, log all events, manage errors, recover from failures, reconcile outputs with source systems
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

133

ETL Tools
Provides facility to specify a large number of transformation rules with a GUI Generate programs to transform data Handle multiple data sources Handle data redundancy Generate metadata as output Most tools exploit parallelism by running on multiple low-cost servers in multi-threaded environment ETL Tools - Second-Generation PowerCentre/Mart from Informatica Data Mart Solution from Sagent Technology DataStage from Ascential
134

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Metadata Management

135

2009 Wipro Ltd - Confidential

What Is Metadata?
Metadata is Information...

That describes the WHAT, WHEN, WHO, WHERE, HOW of the data warehouse About the data being captured and loaded into the Warehouse Documented in IT tools that improves both business and technical understanding of data and data-related processes

136

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Importance Of Metadata
Locating Information Time spent in looking for information. How often information is found? What poor decisions were made based on the incomplete information?

How much money was lost or earned as a result? Interpreting information


How many times have businesses needed to rework or recall products? What impact does it have on the bottom line ? How many mistakes were due to misinterpretation of existing How much interpretation results form too much metadata? How much time is spent trying to determine if any of the metadata is accurate? Integrating information How various data perspectives connect together? How much time is spent trying to figure out that? How much does the inefficiency and lack of metadata affect decision making documentation?

137

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Requirements for DW Metadata Management


Provide a simple catalogue of business metadata descriptions and views Document/manage metadata descriptions from an integrated development environment Enable DW users to identify and invoke pre-built queries against the data stores Design and enhance new data models and schemas for the data warehouse Capture data transformation rules between the operational and data warehousing databases Provide change impact analysis, and update across these technologies
138
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Consumers of Metadata
Technical Users Warehouse administrator Application developer Business Users -Business metadata Meanings Definitions Business Rules Software Tools Used in DW life-cycle development Metadata requirements for each tool must be identified The tool-specific metadata should be analysed for inclusion in the enterprise metadata repository Previously captured metadata should be electronically transferred from the enterprise metadata repository to each individual tool

139

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Trends in the Metadata Management Tools


Third Party Bridging Tools Oracle Exchange
Technology of choice for a long list of repository, enterprise and workgroup vendors

Reischmann-Informatik-Toolbus
Features include facilitation of selective bridging of metadata

Ardent Software/ Dovetail Software -Interplay


Hub and Spoke solution for enabling metadata interoperability Ardent focussing on own engagements, not selling it as independent product

Informix's Metadata Plug-ins


Available with Ardent Datastage version 3.6.2 free of cost for Erwin, Oracle Designer, Sybase Powerdesigner, Brio, Microstrategy
140
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Trends in the Metadata Management Tools


Metadata Repositories IBM, Oracle and Microsoft to offer free or near-free basic repository services Enable organisations to reuse metadata across technologies Integrate DB design, data transformation and BI tools from different vendors Multi-tool vendors taking a bridged or federated rather than integrated approach to sharing metadata Both IBM and Oracle have multiple repositories for different lines of products e.g., One for AD and one for DW, with bridges between them

141

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Trends in the Metadata Management Tools


Metadata Interchange Standards CDIF (CASE Data Interchange Format)
Most frequently used interchange standard Addresses only a limited subset of metadata artifacts

OMG (Object Management Group)-CWM


XML-addresses context and data meaning, not presentation Can enable exchange over the web employing industry standards for storing and sharing programming data Will allow sharing of UML and MOF objects b/w various development tools and repositories

MDC (Metadata Coalition)


Based on XML/UML standards Promoted by Microsoft Along With 20 partners including Object Management Group (OMG), Oracle Carleton Group, CA-PLATINUM Technology (Founding Member), Viasoft
142
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

OLAP

143

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Agenda
OLAP Definition Distinction between OLTP and OLAP

MDDB Concepts
Implementation Techniques Architectures

Features
Representative Tools

6/19/2012

144

144

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

OLAP: On-Line Analytical Processing


OLAP can be defined as a technology which allows the users to view the aggregate data across measurements (like Maturity Amount, Interest Rate etc.) along with a set of related parameters called dimensions (like Product, Organization, Customer, etc.) Used interchangeably with BI Multidimensional view of data is the foundation of OLAP Users :Analysts, Decision makers
6/19/2012 145

145

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Distinction between OLTP and OLAP


OLTP System Source of data Operational data; OLTPs are the original source of the data To control and run fundamental business tasks A snapshot of ongoing business processes Short and fast inserts and updates initiated by end users
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

OLAP System Consolidation data; OLAP data comes from the various OLTP databases Decision support

Purpose of data

What the data reveals Inserts and Updates


6/19/2012

Multi-dimensional views of various kinds of business activities Periodic long-running batch jobs refresh the 146 data

146

MDDB Concepts
A multidimensional database is a computer software system designed to allow for efficient and convenient storage and retrieval of data that is intimately related and stored, viewed and analyzed from different perspectives (Dimensions). A hypercube represents a collection of multidimensional data. The edges of the cube are called dimensions Individual items within each dimensions are called members
147
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

RDBMS v/s MDDB: Increased Complexity...


Relational DBMS
MODEL MINI VAN MINI VAN MINI VAN MINI VAN MINI VAN MINI VAN MINI VAN MINI VAN MINI VAN SPORTS COUPE SPORTS COUPE SPORTS COUPE SPORTS COUPE SPORTS COUPE SPORTS COUPE SPORTS COUPE SPORTS COUPE SPORTS COUPE SEDAN SEDAN SEDAN ... COLOR BLUE BLUE BLUE RED RED RED WHITE WHITE WHITE BLUE BLUE BLUE RED RED RED WHITE WHITE WHITE BLUE BLUE BLUE DEALER Clyde Gleason Carr Clyde Gleason Carr Clyde Gleason Carr Clyde Gleason Carr Clyde Gleason Carr Clyde Gleason Carr Clyde Gleason Carr VOL. 6 3 2 5 3 1 3 1 4 3 3 3 4 3 6 2 3 5 4 3 2 ...

MDDB

Sales Volumes

M O D E L

Mini Van

Coupe Carr Gleason Clyde Blue Red White

Sedan

DEALERSHIP

COLOR

27 x 4 = 108 cells
148

3 x 3 x 3 = 27 cells

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Benefits of MDDB over RDBMS


Ease of Data Presentation & Navigation A great deal of information is gleaned immediately upon direct inspection of the array User is able to view data along presorted dimensions with data arranged in an inherently more organized, and accessible fashion than the one offered by the relational table. Storage Space Very low Space Consumption compared to Relational DB Performance Gives much better performance. Relational DB may give comparable results only through database tuning (indexing, keys etc), which may not be possible for ad-hoc queries. Ease of Maintenance No overhead as data is stored in the same way it is viewed. In Relational DB, indexes, sophisticated joins etc. are used which require considerable storage and maintenance
6/19/2012
149
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

149

Issues with MDDB

Sparsity
- Input data in applications are typically sparse -Increases with increased dimensions

Data Explosion
-Due to Sparsity -Due to Summarization

Performance
-Doesnt perform better than RDBMS at high data volumes (>20-30 GB)

6/19/2012
150
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

150

Issues with MDDB - Sparsity Example


If dimension members of different dimensions do not interact , then blank cell is left behind.
Employee Age
21 19 63 31 27 56 45 41 19
31 41 23 01 14 54 03 12 33

LAST NAME EMP# AGE SMITH 01 21 REGAN 12 Sales Volumes 19 FOX 31 63 Miini Van WELD 14 6 5 31 4 M O KELLY 54 3 5 27 Coupe D 5 E L LINK 03 56 4 3 2 Sedan KRANZ 41 45 Blue Red White LUCUS 33 COLOR 41 WEISS 23 19

Smith

Regan

Fox

L A S T N A M E

Weld

Kelly

Link

Kranz

Lucas

Weiss

EMPLOYEE #

6/19/2012
151
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

151

OLAP Features
Calculations applied across dimensions, through hierarchies and/or across members Trend analysis over sequential time periods, What-if scenarios. Slicing / Dicing subsets for on-screen viewing Rotation to new dimensional comparisons in the viewing area Drill-down/up along the hierarchy Reach-through / Drill-through to underlying detail data

6/19/2012
152
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

152

Features of OLAP - Rotation

Complex Queries & Sorts in Relational environment translated to simple rotation.


Sales Volumes

M O D E L

Mini Van

6 3 4
Blue

5 5 3
Red

4 5 2
White

Coupe

C O L O R ( ROTATE 90 )
o

Blue

6 5 4

3 5 5
MODEL

4 3 2
Sedan

Red

Sedan

White

Mini Van Coupe

COLOR

View #1

View #2

2 dimensional array has 2 views.


6/19/2012
153
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

153

Features of OLAP - Rotation


Sales Volumes

M O D E L

Mini Van Coupe Carr Gleason Clyde Blue Red White

Sedan

C O L O R

Blue

Red White Sedan Coupe Mini Van Carr Gleason Clyde

C O L O R

Blue

Red White Carr Gleason Clyde Mini Van Coupe Sedan

COLOR

( ROTATE 90 )

MODEL

( ROTATE 90 )

DEALERSHIP

( ROTATE 90 )

DEALERSHIP

DEALERSHIP

MODEL

View #1
D E A L E R S H I P D E A L E R S H I P

View #2

View #3

Carr Gleason Mini Van Coupe Sedan White Red Blue

Carr Gleason Blue Red White

Mini Van

Clyde

Clyde Mini Van Coupe Sedan

M O D E L

Coupe Blue Red White Clyde Gleason Carr

Sedan

COLOR

( ROTATE 90 )

MODEL

( ROTATE 90 )

DEALERSHIP

MODEL

COLOR

COLOR

View #4

View #5

View #6

3 dimensional array has 6 views.


6/19/2012
154
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

154

Features of OLAP - Slicing / Filtering


MDDB allows end user to quickly slice in on exact view of the data required.

Sales Volumes

M O D E L

Mini Van Mini Van

Coupe

Coupe Normal Metal Blue Blue

Carr Clyde

Carr Clyde

Normal Blue

Metal Blue

DEALERSHIP

COLOR
6/19/2012
155
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

155

Features of OLAP - Drill Down / Up

ORGANIZATION DIMENSION
REGION Midwest

DISTRICT

Chicago

St. Louis

Gary

DEALERSHIP

Clyde

Gleason

Carr

Levi

Lucas

Bolton

Sales at region/District/Dealership Level

Moving Up and moving down in a hierarchy is referred to as drill-up / roll-up and drill-down

6/19/2012
156
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

156

OLAP Reporting - Drill Down

Inflows ( Region , Year)


200 150 Inflows 100 ($M) 50 0 Year Year 1999 2000 Years

East West Central

6/19/2012
157
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

157

OLAP Reporting - Drill Down

Inflows ( Region , Year - Year 1999)


90 80 70 60 50 Inflows ( $M) 40 30 20 10 0

East West Central

1st Qtr

2nd Qtr 3rd Qtr Year 1999

4th Qtr

Drill-down from Year to Quarter


6/19/2012
158
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

158

OLAP Reporting - Drill Down

Inflows ( Region , Year - Year 1999 - 1st Qtr)


20 15 Inflows ( $M 10 ) 5

East West Central


January February March Year 1999

Drill-down from Quarter to Month

159

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Implementation Techniques -OLAP Architectures

MOLAP - Multidimensional OLAP


Multidimensional Databases for database and application logic layer

ROLAP - Relational OLAP


Access Data stored in relational Data Warehouse for OLAP Analysis. Database and Application logic provided as separate layers

HOLAP - Hybrid OLAP


OLAP Server routes queries first to MDDB, then to RDBMS and result processed on-the-fly in Server

DOLAP - Desk OLAP


Personal MDDB Server and application on the desktop

6/19/2012
160
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

160

MOLAP - MDDB storage

OLAP
Cube
OLAP Calculation Engine

Web Browser

OLAP Tools

OLAP Appli cations


6/19/2012
161
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

161

MOLAP - Features

Powerful analytical capabilities (e.g., financial, forecasting, statistical) Aggregation and calculation capabilities Read/write analytic applications Specialized data structures for
Maximum query performance. Optimum space utilization.
6/19/2012
162
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

162

ROLAP - Standard SQL storage

MDDB - Relational Mapping

Relational DW

Web Browser
OLAP Calculation Engine

SQL

OLAP Tools

OLAP Applications
6/19/2012
163
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

163

ROLAP - Features Three-tier hardware/software architecture:


GUI on client; multidimensional processing on midtier server; target database on database server Processing split between mid-tier & database servers

Ad hoc query capabilities to very large databases DW integration Data scalability

6/19/2012
164
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

164

HOLAP - Combination of RDBMS and MDDB


OLAP Cube
Any Client

Relational DW

Web Browser
OLAP Calculation Engine

SQL

OLAP Tools

OLAP Applications
6/19/2012
165
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

165

HOLAP - Features

RDBMS used for detailed data stored in large databases MDDB used for fast, read/write OLAP analysis and calculations Scalability of RDBMS and MDDB performance Calculation engine provides full analysis features Source of data transparent to end user

6/19/2012
166
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

166

Architecture Comparison

MOLAP
Definition

ROLAP

HOLAP
Hybrid OLAP = ROLAP + summary in MDDB Sparsity exists only in MDDB part To the necessary extent

MDDB OLAP = Relational OLAP = Transaction level data + Transaction level data + summary in MDDB summary in RDBMS Good Design 3 10 times High (May go beyond control. Estimation is very important) Fast - (Depends upon the size of the MDDB) No Sparsity To the necessary extent

Data explosion due to Sparsity Data explosion due to Summarization Query Execution Speed

Slow

Optimum - If the data is fetched from RDBMS then its like ROLAP otherwise like MOLAP. High: RDBMS + disk space + MDDB Server cost Large transactional data + frequent summary analysis

Cost

Medium: MDDB Server + large disk space cost

Low: Only RDBMS + disk space cost

Where to apply?

Small transactional Very large transactional data + complex model + data & it needs to be frequent summary viewed / sorted analysis

6/19/2012
167
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

167

Representative OLAP Tools:

Oracle Express Products Hyperion Essbase Cognos -PowerPlay Seagate - Holos SAS

Micro Strategy - DSS Agent Informix MetaCube Brio Query Business Objects / Web Intelligence

6/19/2012
168
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

168

Sample OLAP Applications

Sales Analysis Financial Analysis Profitability Analysis Performance Analysis Risk Management Profiling & Segmentation Scorecard Application NPA Management Strategic Planning Customer Relationship Management (CRM)
6/19/2012
169
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

169

Data Warehouse Testing

170

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Warehouse Testing Overview


There is an exponentially increasing cost associated with finding software defects later in the development lifecycle. In data warehousing, this is compounded because of the additional business costs of using incorrect data to make critical business decisions

The methodology required for testing a Data Warehouse is different from testing a typical transaction system

171

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Difference In Testing Data warehouse and Transaction System


Data warehouse testing is different on the following counts: User-Triggered vs. System triggered Volume of Test Data Possible scenarios/ Test Cases Programming for testing challenge

172

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Difference In Testing Data warehouse and Transaction System.


User-Triggered vs. System triggered

In data Warehouse, most of the testing is system triggered. Most of the production/Source system testing is the processing of individual transactions, which are driven by some input from the users (Application Form, Servicing Request.). There are very few test cycles, which cover the system-triggered scenarios (Like billing, Valuation.)

173

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Difference In Testing Data warehouse and Transaction System


Volume of Test Data The test data in a transaction system is a very small sample of the overall production data. Data Warehouse has typically large test data as one does try to fill-up maximum possible combination of dimensions and facts. Possible scenarios/ Test Cases In case of Data Warehouse, the permutations and combinations one can possibly test is virtually unlimited due to the core objective of Data Warehouse is to allow all possible views of data.

174

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Difference In Testing Data warehouse and Transaction System


Programming for testing challenge In case of transaction systems, users/business analysts typically test the output of the system. In case of data warehouse, most of the 'Data Warehouse data Quality testing' and ETL testing is done at backend by running separate stand-alone scripts. These scripts compare preTransformation to post Transformation of data.

175

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Warehouse Testing Process


Data-Warehouse testing is basically divided into two parts : 'Back-end' testing where the source systems data is compared to the endresult data in Loaded area 'Front-end' testing where the user checks the data by comparing their MIS with the data displayed by the end-user tools like OLAP. Testing phases consists of : Requirements testing Unit testing Integration testing Performance testing Acceptance testing

176

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Requirements testing
The main aim for doing Requirements testing is to check stated requirements for completeness. Requirements can be tested on following factors. Are the requirements Complete? Are the requirements Singular? Are the requirements Ambiguous? Are the requirements Developable? Are the requirements Testable?

177

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Unit Testing
Unit testing for data warehouses is WHITEBOX. It should check the ETL procedures/mappings/jobs and the reports developed. Unit testing the ETL procedures: Whether ETLs are accessing and picking up right data from right source.

All the data transformations are correct according to the business rules and data warehouse is correctly populated with the transformed data.
Testing the rejected records that dont fulfil transformation rules.

178

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Unit Testing
Unit Testing the Report data:

Verify Report data with source: Data present in a data warehouse will be stored at an aggregate level compare to source systems. QA team should verify the granular data stored in data warehouse against the source data available Field level data verification: QA team must understand the linkages for the fields displayed in the report and should trace back and compare that with the source systems Derivation formulae/calculation rules should be verified

179

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Integration Testing
Integration testing will involve following:

Sequence of ETLs jobs in batch. Initial loading of records on data warehouse. Incremental loading of records at a later date to verify the newly inserted or updated data. Testing the rejected records that dont fulfil transformation rules. Error log generation

180

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Performance Testing
Performance Testing should check for : ETL processes completing within time window.

Monitoring and measuring the data quality issues.


Refresh times for standard/complex reports.

181

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Acceptance testing
Here the system is tested with full functionality and is expected to function as in production. At the end of UAT, the system should be acceptable to the client for use in terms of ETL process integrity and business functionality and reporting.

182

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Questions

183

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Thank You

184

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Components of Warehouse
Source Tables: These are real-time, volatile data in relational databases for transaction processing (OLTP). These can be any relational databases or flat files. ETL Tools: To extract, cleansing, transform (aggregates, joins) and load the data from sources to target. Maintenance and Administration Tools: To authorize and monitor access to the data, set-up users. Scheduling jobs to run on offshore periods. Modeling Tools: Used for data warehouse design for high-performance using dimensional data modeling technique, mapping the source and target files. Databases: Target databases and data marts, which are part of data warehouse. These are structured for analysis and reporting purposes. End-user tools for analysis and reporting: get the reports and analyze the data from target tables. Different types of Querying, Data Mining, OLAP tools are used for this purpose.

185

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Warehouse Architecture


This is a basic design, where there are source files, which are loaded to a warehouse and users query the data for different purposes.

This has a staging area, where the data after cleansing, transforming is loaded and tested here. Later is directly loaded to the target database/warehouse. Which is divided to data marts and can be accessed by different users for their reporting and analyzing purposes.

186

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Modeling
Effective way of using a Data Warehouse

187

2009 Wipro Ltd - Confidential

Data Modeling
Commonly E-R Data Model is used in OLTP, In OLAP Dimensional Data Model is used commonly. E-R (Entity-Relationship) Data Model
Entity: Object that can be observed and classified based on its properties and characteristics. Like employee, book, student Relationship: relating entities to other entities.

Different Perceptive of Data Modeling.


o Conceptual Data Model o Logical Data Model o Physical Data Model

Types of Dimensional Data Models most commonly used:


o Star Schema o Snowflake Schema
188
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Terms used in Dimensional Data Model


To understand dimensional data modeling, let's define some of the terms commonly used in this type of modeling: Dimension: A category of information. For example, the time dimension. Attribute: A unique level within a dimension. For example, Month is an attribute in the Time Dimension. Hierarchy: The specification of levels that represents relationship between different attributes within a dimension. For example, one possible hierarchy in the Time dimension is Year Quarter Month Day. Fact Table: A table that contains the measures of interest. Lookup Table: It provides the detailed information about the attributes. For example, the lookup table for the Quarter attribute would include a list of all of the quarters available in the data warehouse. Surrogate Keys: To avoid the data integrity, surrogate keys are used. They are helpful for Slow Changing Dimensions and act as index/primary keys.
A dimensional model includes fact tables and lookup tables. Fact tables connect to one or more lookup tables, but fact tables do not have direct relationships to one another. Dimensions and hierarchies are represented by lookup tables. Attributes are the non-key columns in the lookup tables.
189
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Star Schema
Dimension Table
product prodId p1 p2 name price bolt 10 nut 5

Dimension Table
store storeId c1 c2 c3 city nyc sfo la

Fact Table
sale oderId date o100 1/7/97 o102 2/7/97 105 3/8/97 custId 53 53 111 prodId p1 p2 p1 storeId c1 c1 c3 qty 1 2 5 amt 12 11 50

Dimension Table
customer custId 53 81 111 name joe fred sally address 10 main 12 main 80 willow city sfo sfo la

190

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Snowflake Schema
Dimension Table Fact Table
store storeId s5 s7 s9 cityId sfo sfo la tId t1 t2 t1 mgr joe fred nancy
sType tId t1 t2 city size small large location downtown suburbs regId north south

Dimension Table
cityId pop sfo 1M la 5M

The star and snowflake schema are most commonly found in dimensional data warehouses and data marts where speed of data retrieval is more important than the efficiency of data manipulations. As such, the tables in these schema are not normalized much, and are frequently designed at a level of normalization short of third normal form.

region regId name north cold region south warm region

191

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Overview of Data Cleansing

192

2009 Wipro Ltd - Confidential

The Need For Data Quality


Difficulty in decision making Time delays in operation Organizational mistrust Data ownership conflicts Customer attrition Costs associated with error detection error rework customer service fixing customer problems

193

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Six Steps To Data Quality


Understand Information Flow In Organization
Identify authoritative data sources Interview Employees & Customers

Identify Potential Problem Areas & Asses Impact

Data Entry Points


Cost of bad data

Measure Quality Of Data

Use business rule discovery tools to identify data with

inconsistent, missing, incomplete, duplicate or incorrect values


Use data cleansing tools to clean data at the source Load only clean data into the data warehouse

Clean & Load Data

Continuous Monitoring

Schedule Periodic Cleansing of Source Data

Identify Areas of Improvement

Identify & Correct Cause of Defects Refine data capture mechanisms at source Educate users on importance of DQ
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

194

Data Quality Solution


Customized Programs Strengths: Addresses specific needs No bulky one time investment Limitations Tons of Custom programs in different environments are difficult to manage Minor alterations demand coding efforts Data Quality Assessment tools Strength Provide automated assessment Limitation No measure of data accuracy

195

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Quality Solution


Business Rule Discovery tools Strengths Detect Correlation in data values Can detect Patterns of behavior that indicate fraud Limitations Not all variables can be discovered Some discovered rules might not be pertinent There may be performance problems with large files or with many fields. Data Reengineering & Cleansing tools Strengths Usually are integrated packages with cleansing features as Addon Limitations Error prevention at source is usually absent The ETL tools have limited cleansing facilities
196
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Tools In The Market


Business Rule Discovery Tools Integrity Data Reengineering Tool from Vality Technology Trillium Software System from Harte -Hanks Data Technologies Migration Architect from DB Star Data Reengineering & Cleansing Tools Carlton Pureview from Oracle ETI-Extract from Evolutionary Technologies PowerMart from Informatica Corp Sagent Data Mart from Sagent Technology Data Quality Assessment Tools Migration Architect, Evoke Axio from Evoke Software Wizrule from Wizsoft Name & Address Cleansing Tools Centrus Suite from Sagent I.d.centric from First Logic

197

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Extraction, Transformation, Load

198

2009 Wipro Ltd - Confidential

ETL Architecture

Visitors

Web Browsers

The Internet

External Data Demographics, Household, Webographics, Income

Staging Area
Web Server Logs & E-comm Transaction Data Flat Files

Meta Data Repository

Scheduled Extraction

RDBMS

Clean Transform Match Merge

Scheduled Loading

Enterprise Data Warehouse

Other OLTP Systems

Data Collection

Data Extraction

Data Transformation

Data Loading

Data Storage & Integration

199

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

ETL Architecture
Data Extraction:
Rummages through a file or database Uses some criteria for selection Identifies qualified data and Transports the data over onto another file or database

Data transformation
Integrating dissimilar data types Changing codes Adding a time attribute Summarizing data Calculating derived values Renormalizing data

Data Extraction Cleanup


Restructuring of records or fields Removal of Operational-only data Supply of missing field values Data Integrity checks Data Consistency and Range checks, etc...

Data loading
Initial and incremental loading Updation of metadata

200

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Why ETL ?
Companies have valuable data lying around throughout their networks that needs to be moved from one place to another. The data lies in all sorts of heterogeneous systems,and therefore in all sorts of formats.

To solve the problem, companies use extract, transform and load (ETL) software.
The data used in ETL processes can come from any source: a mainframe application, an ERP application, a CRM tool, a flat file, and an Excel spreadsheet.

201

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Major components involved in ETL Processing

202

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Major components involved in ETL Processing


Design manager Lets developers define source-to-target mappings, transformations, process flows, and jobs Meta data management Provides a repository to define, document, and manage information about the ETL design and runtime processes Extract The process of reading data from a database. Transform The process of converting the extracted data Load The process of writing the data into the target database. Transport services ETL tools use network and file protocols to move data between source and target systems and in-memory protocols to move data between ETL run-time components. Administration and operation ETL utilities let administrators schedule, run, monitor ETL jobs, log all events, manage errors, recover from failures, reconcile outputs with source systems
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

203

ETL Tools
Provides facility to specify a large number of transformation rules with a GUI Generate programs to transform data Handle multiple data sources Handle data redundancy Generate metadata as output Most tools exploit parallelism by running on multiple low-cost servers in multi-threaded environment ETL Tools - Second-Generation PowerCentre/Mart from Informatica Data Mart Solution from Sagent Technology DataStage from Ascential
204

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Metadata Management

205

2009 Wipro Ltd - Confidential

What Is Metadata?
Metadata is Information...

That describes the WHAT, WHEN, WHO, WHERE, HOW of the data warehouse About the data being captured and loaded into the Warehouse Documented in IT tools that improves both business and technical understanding of data and data-related processes

206

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Importance Of Metadata
Locating Information Time spent in looking for information. How often information is found? What poor decisions were made based on the incomplete information?

How much money was lost or earned as a result? Interpreting information


How many times have businesses needed to rework or recall products? What impact does it have on the bottom line ? How many mistakes were due to misinterpretation of existing How much interpretation results form too much metadata? How much time is spent trying to determine if any of the metadata is accurate? Integrating information How various data perspectives connect together? How much time is spent trying to figure out that? How much does the inefficiency and lack of metadata affect decision making documentation?

207

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Requirements for DW Metadata Management


Provide a simple catalogue of business metadata descriptions and views Document/manage metadata descriptions from an integrated development environment Enable DW users to identify and invoke pre-built queries against the data stores Design and enhance new data models and schemas for the data warehouse Capture data transformation rules between the operational and data warehousing databases Provide change impact analysis, and update across these technologies
208
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Consumers of Metadata
Technical Users Warehouse administrator Application developer Business Users -Business metadata Meanings Definitions Business Rules Software Tools Used in DW life-cycle development Metadata requirements for each tool must be identified The tool-specific metadata should be analysed for inclusion in the enterprise metadata repository Previously captured metadata should be electronically transferred from the enterprise metadata repository to each individual tool

209

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Trends in the Metadata Management Tools


Third Party Bridging Tools Oracle Exchange
Technology of choice for a long list of repository, enterprise and workgroup vendors

Reischmann-Informatik-Toolbus
Features include facilitation of selective bridging of metadata

Ardent Software/ Dovetail Software -Interplay


Hub and Spoke solution for enabling metadata interoperability Ardent focussing on own engagements, not selling it as independent product

Informix's Metadata Plug-ins


Available with Ardent Datastage version 3.6.2 free of cost for Erwin, Oracle Designer, Sybase Powerdesigner, Brio, Microstrategy
210
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Trends in the Metadata Management Tools


Metadata Repositories IBM, Oracle and Microsoft to offer free or near-free basic repository services Enable organisations to reuse metadata across technologies Integrate DB design, data transformation and BI tools from different vendors Multi-tool vendors taking a bridged or federated rather than integrated approach to sharing metadata Both IBM and Oracle have multiple repositories for different lines of products e.g., One for AD and one for DW, with bridges between them

211

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Warehouse Concepts

Avinash Kanumuru Diya Jana Debyajit Majumder

2009 Wipro Ltd - Confidential

Content
1 An Overview of Data Warehouse 2 Data Warehouse Architecture 3 Data Modeling for Data Warehouse 4 Overview of Data Cleansing

5 Data Extraction, Transformation, Load

213

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Warehouse Concepts

Avinash Kanumuru Diya Jana Debyajit Majumder

2009 Wipro Ltd - Confidential

Content
1 An Overview of Data Warehouse 2 Data Warehouse Architecture 3 Data Modeling for Data Warehouse 4 Overview of Data Cleansing

5 Data Extraction, Transformation, Load

215

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Content [contd]
6 Metadata Management 7 OLAP 8 Data Warehouse Testing

216

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

An Overview
Understanding What is a Data Warehouse

217

2009 Wipro Ltd - Confidential

What is Data Warehouse?


Definitions of Data Warehouse A data warehouse is a subject-oriented, integrated, nonvolatile, time-variant collection of data in support of management's decisions. WH Inmon Data Warehouse is a repository of data summarized or aggregated in simplified form from operational systems. End user orientated data access and reporting tools let user get at the data for decision support Babcock A data warehouse is a relational database a copy of transaction data specifically structured for query and analysis Ralph Kimball In simple: Data warehousing is collection of data from different systems, which helps in Business Decisions, Analysis and Reporting.

218

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Warehouse def. by WH Inmon


A common way of introducing data warehousing is to refer to the characteristics of a data warehouse as set forth by William Inmon: Subject Oriented Data that gives information about a particular subject instead of about a company's ongoing operations. Integrated Data that is gathered into the data warehouse from a variety of sources and merged into a coherent whole. Nonvolatile Data is stable in a data warehouse. More data is added but data is never removed. This enables management to gain a consistent picture of the business. Time Variant In order to discover trends in business, analysts need large amounts of data. This is very much in contrast to online transaction processing (OLTP) systems, where performance requirements demand that historical data be moved to an archive. All data in the data warehouse is identified with a particular time period.

219

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Warehouse Architecture


What makes a Data Warehouse

220

2009 Wipro Ltd - Confidential

Components of Warehouse
Source Tables: These are real-time, volatile data in relational databases for transaction processing (OLTP). These can be any relational databases or flat files. ETL Tools: To extract, cleansing, transform (aggregates, joins) and load the data from sources to target. Maintenance and Administration Tools: To authorize and monitor access to the data, set-up users. Scheduling jobs to run on offshore periods. Modeling Tools: Used for data warehouse design for high-performance using dimensional data modeling technique, mapping the source and target files. Databases: Target databases and data marts, which are part of data warehouse. These are structured for analysis and reporting purposes. End-user tools for analysis and reporting: get the reports and analyze the data from target tables. Different types of Querying, Data Mining, OLAP tools are used for this purpose.

221

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Warehouse Architecture


This is a basic design, where there are source files, which are loaded to a warehouse and users query the data for different purposes.

This has a staging area, where the data after cleansing, transforming is loaded and tested here. Later is directly loaded to the target database/warehouse. Which is divided to data marts and can be accessed by different users for their reporting and analyzing purposes.

222

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Modeling
Effective way of using a Data Warehouse

223

2009 Wipro Ltd - Confidential

Data Modeling
Commonly E-R Data Model is used in OLTP, In OLAP Dimensional Data Model is used commonly. E-R (Entity-Relationship) Data Model
Entity: Object that can be observed and classified based on its properties and characteristics. Like employee, book, student Relationship: relating entities to other entities.

Different Perceptive of Data Modeling.


o Conceptual Data Model o Logical Data Model o Physical Data Model

Types of Dimensional Data Models most commonly used:


o Star Schema o Snowflake Schema
224
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Terms used in Dimensional Data Model


To understand dimensional data modeling, let's define some of the terms commonly used in this type of modeling: Dimension: A category of information. For example, the time dimension. Attribute: A unique level within a dimension. For example, Month is an attribute in the Time Dimension. Hierarchy: The specification of levels that represents relationship between different attributes within a dimension. For example, one possible hierarchy in the Time dimension is Year Quarter Month Day. Fact Table: A table that contains the measures of interest. Lookup Table: It provides the detailed information about the attributes. For example, the lookup table for the Quarter attribute would include a list of all of the quarters available in the data warehouse. Surrogate Keys: To avoid the data integrity, surrogate keys are used. They are helpful for Slow Changing Dimensions and act as index/primary keys.
A dimensional model includes fact tables and lookup tables. Fact tables connect to one or more lookup tables, but fact tables do not have direct relationships to one another. Dimensions and hierarchies are represented by lookup tables. Attributes are the non-key columns in the lookup tables.
225
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Star Schema
Dimension Table
product prodId p1 p2 name price bolt 10 nut 5

Dimension Table
store storeId c1 c2 c3 city nyc sfo la

Fact Table
sale oderId date o100 1/7/97 o102 2/7/97 105 3/8/97 custId 53 53 111 prodId p1 p2 p1 storeId c1 c1 c3 qty 1 2 5 amt 12 11 50

Dimension Table
customer custId 53 81 111 name joe fred sally address 10 main 12 main 80 willow city sfo sfo la

226

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Snowflake Schema
Dimension Table Fact Table
store storeId s5 s7 s9 cityId sfo sfo la tId t1 t2 t1 mgr joe fred nancy
sType tId t1 t2 city size small large location downtown suburbs regId north south

Dimension Table
cityId pop sfo 1M la 5M

The star and snowflake schema are most commonly found in dimensional data warehouses and data marts where speed of data retrieval is more important than the efficiency of data manipulations. As such, the tables in these schema are not normalized much, and are frequently designed at a level of normalization short of third normal form.

region regId name north cold region south warm region

227

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Overview of Data Cleansing

228

2009 Wipro Ltd - Confidential

The Need For Data Quality


Difficulty in decision making Time delays in operation Organizational mistrust Data ownership conflicts Customer attrition Costs associated with error detection error rework customer service fixing customer problems

229

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Six Steps To Data Quality


Understand Information Flow In Organization
Identify authoritative data sources Interview Employees & Customers

Identify Potential Problem Areas & Asses Impact

Data Entry Points


Cost of bad data

Measure Quality Of Data

Use business rule discovery tools to identify data with

inconsistent, missing, incomplete, duplicate or incorrect values


Use data cleansing tools to clean data at the source Load only clean data into the data warehouse

Clean & Load Data

Continuous Monitoring

Schedule Periodic Cleansing of Source Data

Identify Areas of Improvement

Identify & Correct Cause of Defects Refine data capture mechanisms at source Educate users on importance of DQ
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

230

Data Quality Solution


Customized Programs Strengths: Addresses specific needs No bulky one time investment Limitations Tons of Custom programs in different environments are difficult to manage Minor alterations demand coding efforts Data Quality Assessment tools Strength Provide automated assessment Limitation No measure of data accuracy

231

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Quality Solution


Business Rule Discovery tools Strengths Detect Correlation in data values Can detect Patterns of behavior that indicate fraud Limitations Not all variables can be discovered Some discovered rules might not be pertinent There may be performance problems with large files or with many fields. Data Reengineering & Cleansing tools Strengths Usually are integrated packages with cleansing features as Addon Limitations Error prevention at source is usually absent The ETL tools have limited cleansing facilities
232
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Tools In The Market


Business Rule Discovery Tools Integrity Data Reengineering Tool from Vality Technology Trillium Software System from Harte -Hanks Data Technologies Migration Architect from DB Star Data Reengineering & Cleansing Tools Carlton Pureview from Oracle ETI-Extract from Evolutionary Technologies PowerMart from Informatica Corp Sagent Data Mart from Sagent Technology Data Quality Assessment Tools Migration Architect, Evoke Axio from Evoke Software Wizrule from Wizsoft Name & Address Cleansing Tools Centrus Suite from Sagent I.d.centric from First Logic

233

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Extraction, Transformation, Load

234

2009 Wipro Ltd - Confidential

ETL Architecture

Visitors

Web Browsers

The Internet

External Data Demographics, Household, Webographics, Income

Staging Area
Web Server Logs & E-comm Transaction Data Flat Files

Meta Data Repository

Scheduled Extraction

RDBMS

Clean Transform Match Merge

Scheduled Loading

Enterprise Data Warehouse

Other OLTP Systems

Data Collection

Data Extraction

Data Transformation

Data Loading

Data Storage & Integration

235

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

ETL Architecture
Data Extraction:
Rummages through a file or database Uses some criteria for selection Identifies qualified data and Transports the data over onto another file or database

Data transformation
Integrating dissimilar data types Changing codes Adding a time attribute Summarizing data Calculating derived values Renormalizing data

Data Extraction Cleanup


Restructuring of records or fields Removal of Operational-only data Supply of missing field values Data Integrity checks Data Consistency and Range checks, etc...

Data loading
Initial and incremental loading Updation of metadata

236

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Why ETL ?
Companies have valuable data lying around throughout their networks that needs to be moved from one place to another. The data lies in all sorts of heterogeneous systems,and therefore in all sorts of formats.

To solve the problem, companies use extract, transform and load (ETL) software.
The data used in ETL processes can come from any source: a mainframe application, an ERP application, a CRM tool, a flat file, and an Excel spreadsheet.

237

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Major components involved in ETL Processing

238

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Major components involved in ETL Processing


Design manager Lets developers define source-to-target mappings, transformations, process flows, and jobs Meta data management Provides a repository to define, document, and manage information about the ETL design and runtime processes Extract The process of reading data from a database. Transform The process of converting the extracted data Load The process of writing the data into the target database. Transport services ETL tools use network and file protocols to move data between source and target systems and in-memory protocols to move data between ETL run-time components. Administration and operation ETL utilities let administrators schedule, run, monitor ETL jobs, log all events, manage errors, recover from failures, reconcile outputs with source systems
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

239

ETL Tools
Provides facility to specify a large number of transformation rules with a GUI Generate programs to transform data Handle multiple data sources Handle data redundancy Generate metadata as output Most tools exploit parallelism by running on multiple low-cost servers in multi-threaded environment ETL Tools - Second-Generation PowerCentre/Mart from Informatica Data Mart Solution from Sagent Technology DataStage from Ascential
240

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Metadata Management

241

2009 Wipro Ltd - Confidential

What Is Metadata?
Metadata is Information...

That describes the WHAT, WHEN, WHO, WHERE, HOW of the data warehouse About the data being captured and loaded into the Warehouse Documented in IT tools that improves both business and technical understanding of data and data-related processes

242

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Importance Of Metadata
Locating Information Time spent in looking for information. How often information is found? What poor decisions were made based on the incomplete information?

How much money was lost or earned as a result? Interpreting information


How many times have businesses needed to rework or recall products? What impact does it have on the bottom line ? How many mistakes were due to misinterpretation of existing How much interpretation results form too much metadata? How much time is spent trying to determine if any of the metadata is accurate? Integrating information How various data perspectives connect together? How much time is spent trying to figure out that? How much does the inefficiency and lack of metadata affect decision making documentation?

243

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Requirements for DW Metadata Management


Provide a simple catalogue of business metadata descriptions and views Document/manage metadata descriptions from an integrated development environment Enable DW users to identify and invoke pre-built queries against the data stores Design and enhance new data models and schemas for the data warehouse Capture data transformation rules between the operational and data warehousing databases Provide change impact analysis, and update across these technologies
244
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Consumers of Metadata
Technical Users Warehouse administrator Application developer Business Users -Business metadata Meanings Definitions Business Rules Software Tools Used in DW life-cycle development Metadata requirements for each tool must be identified The tool-specific metadata should be analysed for inclusion in the enterprise metadata repository Previously captured metadata should be electronically transferred from the enterprise metadata repository to each individual tool

245

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Trends in the Metadata Management Tools


Third Party Bridging Tools Oracle Exchange
Technology of choice for a long list of repository, enterprise and workgroup vendors

Reischmann-Informatik-Toolbus
Features include facilitation of selective bridging of metadata

Ardent Software/ Dovetail Software -Interplay


Hub and Spoke solution for enabling metadata interoperability Ardent focussing on own engagements, not selling it as independent product

Informix's Metadata Plug-ins


Available with Ardent Datastage version 3.6.2 free of cost for Erwin, Oracle Designer, Sybase Powerdesigner, Brio, Microstrategy
246
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Trends in the Metadata Management Tools


Metadata Repositories IBM, Oracle and Microsoft to offer free or near-free basic repository services Enable organisations to reuse metadata across technologies Integrate DB design, data transformation and BI tools from different vendors Multi-tool vendors taking a bridged or federated rather than integrated approach to sharing metadata Both IBM and Oracle have multiple repositories for different lines of products e.g., One for AD and one for DW, with bridges between them

247

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Trends in the Metadata Management Tools


Metadata Interchange Standards CDIF (CASE Data Interchange Format)
Most frequently used interchange standard Addresses only a limited subset of metadata artifacts

OMG (Object Management Group)-CWM


XML-addresses context and data meaning, not presentation Can enable exchange over the web employing industry standards for storing and sharing programming data Will allow sharing of UML and MOF objects b/w various development tools and repositories

MDC (Metadata Coalition)


Based on XML/UML standards Promoted by Microsoft Along With 20 partners including Object Management Group (OMG), Oracle Carleton Group, CA-PLATINUM Technology (Founding Member), Viasoft
248
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

OLAP

249

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Agenda
OLAP Definition Distinction between OLTP and OLAP

MDDB Concepts
Implementation Techniques Architectures

Features
Representative Tools

6/19/2012

250

250

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

OLAP: On-Line Analytical Processing


OLAP can be defined as a technology which allows the users to view the aggregate data across measurements (like Maturity Amount, Interest Rate etc.) along with a set of related parameters called dimensions (like Product, Organization, Customer, etc.) Used interchangeably with BI Multidimensional view of data is the foundation of OLAP Users :Analysts, Decision makers
6/19/2012 251

251

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Distinction between OLTP and OLAP


OLTP System Source of data Operational data; OLTPs are the original source of the data To control and run fundamental business tasks A snapshot of ongoing business processes Short and fast inserts and updates initiated by end users
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

OLAP System Consolidation data; OLAP data comes from the various OLTP databases Decision support

Purpose of data

What the data reveals Inserts and Updates


6/19/2012

Multi-dimensional views of various kinds of business activities Periodic long-running batch jobs refresh the 252 data

252

MDDB Concepts
A multidimensional database is a computer software system designed to allow for efficient and convenient storage and retrieval of data that is intimately related and stored, viewed and analyzed from different perspectives (Dimensions). A hypercube represents a collection of multidimensional data. The edges of the cube are called dimensions Individual items within each dimensions are called members
253
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

RDBMS v/s MDDB: Increased Complexity...


Relational DBMS
MODEL MINI VAN MINI VAN MINI VAN MINI VAN MINI VAN MINI VAN MINI VAN MINI VAN MINI VAN SPORTS COUPE SPORTS COUPE SPORTS COUPE SPORTS COUPE SPORTS COUPE SPORTS COUPE SPORTS COUPE SPORTS COUPE SPORTS COUPE SEDAN SEDAN SEDAN ... COLOR BLUE BLUE BLUE RED RED RED WHITE WHITE WHITE BLUE BLUE BLUE RED RED RED WHITE WHITE WHITE BLUE BLUE BLUE DEALER Clyde Gleason Carr Clyde Gleason Carr Clyde Gleason Carr Clyde Gleason Carr Clyde Gleason Carr Clyde Gleason Carr Clyde Gleason Carr VOL. 6 3 2 5 3 1 3 1 4 3 3 3 4 3 6 2 3 5 4 3 2 ...

MDDB

Sales Volumes

M O D E L

Mini Van

Coupe Carr Gleason Clyde Blue Red White

Sedan

DEALERSHIP

COLOR

27 x 4 = 108 cells
254

3 x 3 x 3 = 27 cells

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Benefits of MDDB over RDBMS


Ease of Data Presentation & Navigation A great deal of information is gleaned immediately upon direct inspection of the array User is able to view data along presorted dimensions with data arranged in an inherently more organized, and accessible fashion than the one offered by the relational table. Storage Space Very low Space Consumption compared to Relational DB Performance Gives much better performance. Relational DB may give comparable results only through database tuning (indexing, keys etc), which may not be possible for ad-hoc queries. Ease of Maintenance No overhead as data is stored in the same way it is viewed. In Relational DB, indexes, sophisticated joins etc. are used which require considerable storage and maintenance
6/19/2012
255
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

255

Issues with MDDB

Sparsity
- Input data in applications are typically sparse -Increases with increased dimensions

Data Explosion
-Due to Sparsity -Due to Summarization

Performance
-Doesnt perform better than RDBMS at high data volumes (>20-30 GB)

6/19/2012
256
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

256

Issues with MDDB - Sparsity Example


If dimension members of different dimensions do not interact , then blank cell is left behind.
Employee Age
21 19 63 31 27 56 45 41 19
31 41 23 01 14 54 03 12 33

LAST NAME EMP# AGE SMITH 01 21 REGAN 12 Sales Volumes 19 FOX 31 63 Miini Van WELD 14 6 5 31 4 M O KELLY 54 3 5 27 Coupe D 5 E L LINK 03 56 4 3 2 Sedan KRANZ 41 45 Blue Red White LUCUS 33 COLOR 41 WEISS 23 19

Smith

Regan

Fox

L A S T N A M E

Weld

Kelly

Link

Kranz

Lucas

Weiss

EMPLOYEE #

6/19/2012
257
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

257

OLAP Features
Calculations applied across dimensions, through hierarchies and/or across members Trend analysis over sequential time periods, What-if scenarios. Slicing / Dicing subsets for on-screen viewing Rotation to new dimensional comparisons in the viewing area Drill-down/up along the hierarchy Reach-through / Drill-through to underlying detail data

6/19/2012
258
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

258

Features of OLAP - Rotation

Complex Queries & Sorts in Relational environment translated to simple rotation.


Sales Volumes

M O D E L

Mini Van

6 3 4
Blue

5 5 3
Red

4 5 2
White

Coupe

C O L O R ( ROTATE 90 )
o

Blue

6 5 4

3 5 5
MODEL

4 3 2
Sedan

Red

Sedan

White

Mini Van Coupe

COLOR

View #1

View #2

2 dimensional array has 2 views.


6/19/2012
259
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

259

Features of OLAP - Rotation


Sales Volumes

M O D E L

Mini Van Coupe Carr Gleason Clyde Blue Red White

Sedan

C O L O R

Blue

Red White Sedan Coupe Mini Van Carr Gleason Clyde

C O L O R

Blue

Red White Carr Gleason Clyde Mini Van Coupe Sedan

COLOR

( ROTATE 90 )

MODEL

( ROTATE 90 )

DEALERSHIP

( ROTATE 90 )

DEALERSHIP

DEALERSHIP

MODEL

View #1
D E A L E R S H I P D E A L E R S H I P

View #2

View #3

Carr Gleason Mini Van Coupe Sedan White Red Blue

Carr Gleason Blue Red White

Mini Van

Clyde

Clyde Mini Van Coupe Sedan

M O D E L

Coupe Blue Red White Clyde Gleason Carr

Sedan

COLOR

( ROTATE 90 )

MODEL

( ROTATE 90 )

DEALERSHIP

MODEL

COLOR

COLOR

View #4

View #5

View #6

3 dimensional array has 6 views.


6/19/2012
260
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

260

Features of OLAP - Slicing / Filtering


MDDB allows end user to quickly slice in on exact view of the data required.

Sales Volumes

M O D E L

Mini Van Mini Van

Coupe

Coupe Normal Metal Blue Blue

Carr Clyde

Carr Clyde

Normal Blue

Metal Blue

DEALERSHIP

COLOR
6/19/2012
261
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

261

Features of OLAP - Drill Down / Up

ORGANIZATION DIMENSION
REGION Midwest

DISTRICT

Chicago

St. Louis

Gary

DEALERSHIP

Clyde

Gleason

Carr

Levi

Lucas

Bolton

Sales at region/District/Dealership Level

Moving Up and moving down in a hierarchy is referred to as drill-up / roll-up and drill-down

6/19/2012
262
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

262

OLAP Reporting - Drill Down

Inflows ( Region , Year)


200 150 Inflows 100 ($M) 50 0 Year Year 1999 2000 Years

East West Central

6/19/2012
263
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

263

OLAP Reporting - Drill Down

Inflows ( Region , Year - Year 1999)


90 80 70 60 50 Inflows ( $M) 40 30 20 10 0

East West Central

1st Qtr

2nd Qtr 3rd Qtr Year 1999

4th Qtr

Drill-down from Year to Quarter


6/19/2012
264
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

264

OLAP Reporting - Drill Down

Inflows ( Region , Year - Year 1999 - 1st Qtr)


20 15 Inflows ( $M 10 ) 5

East West Central


January February March Year 1999

Drill-down from Quarter to Month

265

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Implementation Techniques -OLAP Architectures

MOLAP - Multidimensional OLAP


Multidimensional Databases for database and application logic layer

ROLAP - Relational OLAP


Access Data stored in relational Data Warehouse for OLAP Analysis. Database and Application logic provided as separate layers

HOLAP - Hybrid OLAP


OLAP Server routes queries first to MDDB, then to RDBMS and result processed on-the-fly in Server

DOLAP - Desk OLAP


Personal MDDB Server and application on the desktop

6/19/2012
266
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

266

MOLAP - MDDB storage

OLAP
Cube
OLAP Calculation Engine

Web Browser

OLAP Tools

OLAP Appli cations


6/19/2012
267
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

267

MOLAP - Features

Powerful analytical capabilities (e.g., financial, forecasting, statistical) Aggregation and calculation capabilities Read/write analytic applications Specialized data structures for
Maximum query performance. Optimum space utilization.
6/19/2012
268
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

268

ROLAP - Standard SQL storage

MDDB - Relational Mapping

Relational DW

Web Browser
OLAP Calculation Engine

SQL

OLAP Tools

OLAP Applications
6/19/2012
269
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

269

ROLAP - Features Three-tier hardware/software architecture:


GUI on client; multidimensional processing on midtier server; target database on database server Processing split between mid-tier & database servers

Ad hoc query capabilities to very large databases DW integration Data scalability

6/19/2012
270
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

270

HOLAP - Combination of RDBMS and MDDB


OLAP Cube
Any Client

Relational DW

Web Browser
OLAP Calculation Engine

SQL

OLAP Tools

OLAP Applications
6/19/2012
271
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

271

HOLAP - Features

RDBMS used for detailed data stored in large databases MDDB used for fast, read/write OLAP analysis and calculations Scalability of RDBMS and MDDB performance Calculation engine provides full analysis features Source of data transparent to end user

6/19/2012
272
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

272

Architecture Comparison

MOLAP
Definition

ROLAP

HOLAP
Hybrid OLAP = ROLAP + summary in MDDB Sparsity exists only in MDDB part To the necessary extent

MDDB OLAP = Relational OLAP = Transaction level data + Transaction level data + summary in MDDB summary in RDBMS Good Design 3 10 times High (May go beyond control. Estimation is very important) Fast - (Depends upon the size of the MDDB) No Sparsity To the necessary extent

Data explosion due to Sparsity Data explosion due to Summarization Query Execution Speed

Slow

Optimum - If the data is fetched from RDBMS then its like ROLAP otherwise like MOLAP. High: RDBMS + disk space + MDDB Server cost Large transactional data + frequent summary analysis

Cost

Medium: MDDB Server + large disk space cost

Low: Only RDBMS + disk space cost

Where to apply?

Small transactional Very large transactional data + complex model + data & it needs to be frequent summary viewed / sorted analysis

6/19/2012
273
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

273

Representative OLAP Tools:

Oracle Express Products Hyperion Essbase Cognos -PowerPlay Seagate - Holos SAS

Micro Strategy - DSS Agent Informix MetaCube Brio Query Business Objects / Web Intelligence

6/19/2012
274
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

274

Sample OLAP Applications

Sales Analysis Financial Analysis Profitability Analysis Performance Analysis Risk Management Profiling & Segmentation Scorecard Application NPA Management Strategic Planning Customer Relationship Management (CRM)
6/19/2012
275
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

275

Data Warehouse Testing

276

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Warehouse Testing Overview


There is an exponentially increasing cost associated with finding software defects later in the development lifecycle. In data warehousing, this is compounded because of the additional business costs of using incorrect data to make critical business decisions

The methodology required for testing a Data Warehouse is different from testing a typical transaction system

277

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Difference In Testing Data warehouse and Transaction System


Data warehouse testing is different on the following counts: User-Triggered vs. System triggered Volume of Test Data Possible scenarios/ Test Cases Programming for testing challenge

278

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Difference In Testing Data warehouse and Transaction System.


User-Triggered vs. System triggered

In data Warehouse, most of the testing is system triggered. Most of the production/Source system testing is the processing of individual transactions, which are driven by some input from the users (Application Form, Servicing Request.). There are very few test cycles, which cover the system-triggered scenarios (Like billing, Valuation.)

279

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Difference In Testing Data warehouse and Transaction System


Volume of Test Data The test data in a transaction system is a very small sample of the overall production data. Data Warehouse has typically large test data as one does try to fill-up maximum possible combination of dimensions and facts. Possible scenarios/ Test Cases In case of Data Warehouse, the permutations and combinations one can possibly test is virtually unlimited due to the core objective of Data Warehouse is to allow all possible views of data.

280

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Difference In Testing Data warehouse and Transaction System


Programming for testing challenge In case of transaction systems, users/business analysts typically test the output of the system. In case of data warehouse, most of the 'Data Warehouse data Quality testing' and ETL testing is done at backend by running separate stand-alone scripts. These scripts compare preTransformation to post Transformation of data.

281

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Warehouse Testing Process


Data-Warehouse testing is basically divided into two parts : 'Back-end' testing where the source systems data is compared to the endresult data in Loaded area 'Front-end' testing where the user checks the data by comparing their MIS with the data displayed by the end-user tools like OLAP. Testing phases consists of : Requirements testing Unit testing Integration testing Performance testing Acceptance testing

282

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Requirements testing
The main aim for doing Requirements testing is to check stated requirements for completeness. Requirements can be tested on following factors. Are the requirements Complete? Are the requirements Singular? Are the requirements Ambiguous? Are the requirements Developable? Are the requirements Testable?

283

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Unit Testing
Unit testing for data warehouses is WHITEBOX. It should check the ETL procedures/mappings/jobs and the reports developed. Unit testing the ETL procedures: Whether ETLs are accessing and picking up right data from right source.

All the data transformations are correct according to the business rules and data warehouse is correctly populated with the transformed data.
Testing the rejected records that dont fulfil transformation rules.

284

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Unit Testing
Unit Testing the Report data:

Verify Report data with source: Data present in a data warehouse will be stored at an aggregate level compare to source systems. QA team should verify the granular data stored in data warehouse against the source data available Field level data verification: QA team must understand the linkages for the fields displayed in the report and should trace back and compare that with the source systems Derivation formulae/calculation rules should be verified

285

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Integration Testing
Integration testing will involve following:

Sequence of ETLs jobs in batch. Initial loading of records on data warehouse. Incremental loading of records at a later date to verify the newly inserted or updated data. Testing the rejected records that dont fulfil transformation rules. Error log generation

286

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Performance Testing
Performance Testing should check for : ETL processes completing within time window.

Monitoring and measuring the data quality issues.


Refresh times for standard/complex reports.

287

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Acceptance testing
Here the system is tested with full functionality and is expected to function as in production. At the end of UAT, the system should be acceptable to the client for use in terms of ETL process integrity and business functionality and reporting.

288

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Questions

289

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Thank You

290

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Content [contd]
6 Metadata Management 7 OLAP 8 Data Warehouse Testing

291

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

An Overview
Understanding What is a Data Warehouse

292

2009 Wipro Ltd - Confidential

What is Data Warehouse?


Definitions of Data Warehouse A data warehouse is a subject-oriented, integrated, nonvolatile, time-variant collection of data in support of management's decisions. WH Inmon Data Warehouse is a repository of data summarized or aggregated in simplified form from operational systems. End user orientated data access and reporting tools let user get at the data for decision support Babcock A data warehouse is a relational database a copy of transaction data specifically structured for query and analysis Ralph Kimball In simple: Data warehousing is collection of data from different systems, which helps in Business Decisions, Analysis and Reporting.

293

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Warehouse def. by WH Inmon


A common way of introducing data warehousing is to refer to the characteristics of a data warehouse as set forth by William Inmon: Subject Oriented Data that gives information about a particular subject instead of about a company's ongoing operations. Integrated Data that is gathered into the data warehouse from a variety of sources and merged into a coherent whole. Nonvolatile Data is stable in a data warehouse. More data is added but data is never removed. This enables management to gain a consistent picture of the business. Time Variant In order to discover trends in business, analysts need large amounts of data. This is very much in contrast to online transaction processing (OLTP) systems, where performance requirements demand that historical data be moved to an archive. All data in the data warehouse is identified with a particular time period.

294

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Warehouse Architecture


What makes a Data Warehouse

295

2009 Wipro Ltd - Confidential

Components of Warehouse
Source Tables: These are real-time, volatile data in relational databases for transaction processing (OLTP). These can be any relational databases or flat files. ETL Tools: To extract, cleansing, transform (aggregates, joins) and load the data from sources to target. Maintenance and Administration Tools: To authorize and monitor access to the data, set-up users. Scheduling jobs to run on offshore periods. Modeling Tools: Used for data warehouse design for high-performance using dimensional data modeling technique, mapping the source and target files. Databases: Target databases and data marts, which are part of data warehouse. These are structured for analysis and reporting purposes. End-user tools for analysis and reporting: get the reports and analyze the data from target tables. Different types of Querying, Data Mining, OLAP tools are used for this purpose.

296

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Warehouse Architecture


This is a basic design, where there are source files, which are loaded to a warehouse and users query the data for different purposes.

This has a staging area, where the data after cleansing, transforming is loaded and tested here. Later is directly loaded to the target database/warehouse. Which is divided to data marts and can be accessed by different users for their reporting and analyzing purposes.

297

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Modeling
Effective way of using a Data Warehouse

298

2009 Wipro Ltd - Confidential

Data Modeling
Commonly E-R Data Model is used in OLTP, In OLAP Dimensional Data Model is used commonly. E-R (Entity-Relationship) Data Model
Entity: Object that can be observed and classified based on its properties and characteristics. Like employee, book, student Relationship: relating entities to other entities.

Different Perceptive of Data Modeling.


o Conceptual Data Model o Logical Data Model o Physical Data Model

Types of Dimensional Data Models most commonly used:


o Star Schema o Snowflake Schema
299
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Terms used in Dimensional Data Model


To understand dimensional data modeling, let's define some of the terms commonly used in this type of modeling: Dimension: A category of information. For example, the time dimension. Attribute: A unique level within a dimension. For example, Month is an attribute in the Time Dimension. Hierarchy: The specification of levels that represents relationship between different attributes within a dimension. For example, one possible hierarchy in the Time dimension is Year Quarter Month Day. Fact Table: A table that contains the measures of interest. Lookup Table: It provides the detailed information about the attributes. For example, the lookup table for the Quarter attribute would include a list of all of the quarters available in the data warehouse. Surrogate Keys: To avoid the data integrity, surrogate keys are used. They are helpful for Slow Changing Dimensions and act as index/primary keys.
A dimensional model includes fact tables and lookup tables. Fact tables connect to one or more lookup tables, but fact tables do not have direct relationships to one another. Dimensions and hierarchies are represented by lookup tables. Attributes are the non-key columns in the lookup tables.
300
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Star Schema
Dimension Table
product prodId p1 p2 name price bolt 10 nut 5

Dimension Table
store storeId c1 c2 c3 city nyc sfo la

Fact Table
sale oderId date o100 1/7/97 o102 2/7/97 105 3/8/97 custId 53 53 111 prodId p1 p2 p1 storeId c1 c1 c3 qty 1 2 5 amt 12 11 50

Dimension Table
customer custId 53 81 111 name joe fred sally address 10 main 12 main 80 willow city sfo sfo la

301

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Snowflake Schema
Dimension Table Fact Table
store storeId s5 s7 s9 cityId sfo sfo la tId t1 t2 t1 mgr joe fred nancy
sType tId t1 t2 city size small large location downtown suburbs regId north south

Dimension Table
cityId pop sfo 1M la 5M

The star and snowflake schema are most commonly found in dimensional data warehouses and data marts where speed of data retrieval is more important than the efficiency of data manipulations. As such, the tables in these schema are not normalized much, and are frequently designed at a level of normalization short of third normal form.

region regId name north cold region south warm region

302

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Overview of Data Cleansing

303

2009 Wipro Ltd - Confidential

The Need For Data Quality


Difficulty in decision making Time delays in operation Organizational mistrust Data ownership conflicts Customer attrition Costs associated with error detection error rework customer service fixing customer problems

304

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Six Steps To Data Quality


Understand Information Flow In Organization
Identify authoritative data sources Interview Employees & Customers

Identify Potential Problem Areas & Asses Impact

Data Entry Points


Cost of bad data

Measure Quality Of Data

Use business rule discovery tools to identify data with

inconsistent, missing, incomplete, duplicate or incorrect values


Use data cleansing tools to clean data at the source Load only clean data into the data warehouse

Clean & Load Data

Continuous Monitoring

Schedule Periodic Cleansing of Source Data

Identify Areas of Improvement

Identify & Correct Cause of Defects Refine data capture mechanisms at source Educate users on importance of DQ
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

305

Data Quality Solution


Customized Programs Strengths: Addresses specific needs No bulky one time investment Limitations Tons of Custom programs in different environments are difficult to manage Minor alterations demand coding efforts Data Quality Assessment tools Strength Provide automated assessment Limitation No measure of data accuracy

306

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Quality Solution


Business Rule Discovery tools Strengths Detect Correlation in data values Can detect Patterns of behavior that indicate fraud Limitations Not all variables can be discovered Some discovered rules might not be pertinent There may be performance problems with large files or with many fields. Data Reengineering & Cleansing tools Strengths Usually are integrated packages with cleansing features as Addon Limitations Error prevention at source is usually absent The ETL tools have limited cleansing facilities
307
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Tools In The Market


Business Rule Discovery Tools Integrity Data Reengineering Tool from Vality Technology Trillium Software System from Harte -Hanks Data Technologies Migration Architect from DB Star Data Reengineering & Cleansing Tools Carlton Pureview from Oracle ETI-Extract from Evolutionary Technologies PowerMart from Informatica Corp Sagent Data Mart from Sagent Technology Data Quality Assessment Tools Migration Architect, Evoke Axio from Evoke Software Wizrule from Wizsoft Name & Address Cleansing Tools Centrus Suite from Sagent I.d.centric from First Logic

308

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Extraction, Transformation, Load

309

2009 Wipro Ltd - Confidential

ETL Architecture

Visitors

Web Browsers

The Internet

External Data Demographics, Household, Webographics, Income

Staging Area
Web Server Logs & E-comm Transaction Data Flat Files

Meta Data Repository

Scheduled Extraction

RDBMS

Clean Transform Match Merge

Scheduled Loading

Enterprise Data Warehouse

Other OLTP Systems

Data Collection

Data Extraction

Data Transformation

Data Loading

Data Storage & Integration

310

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

ETL Architecture
Data Extraction:
Rummages through a file or database Uses some criteria for selection Identifies qualified data and Transports the data over onto another file or database

Data transformation
Integrating dissimilar data types Changing codes Adding a time attribute Summarizing data Calculating derived values Renormalizing data

Data Extraction Cleanup


Restructuring of records or fields Removal of Operational-only data Supply of missing field values Data Integrity checks Data Consistency and Range checks, etc...

Data loading
Initial and incremental loading Updation of metadata

311

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Why ETL ?
Companies have valuable data lying around throughout their networks that needs to be moved from one place to another. The data lies in all sorts of heterogeneous systems,and therefore in all sorts of formats.

To solve the problem, companies use extract, transform and load (ETL) software.
The data used in ETL processes can come from any source: a mainframe application, an ERP application, a CRM tool, a flat file, and an Excel spreadsheet.

312

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Major components involved in ETL Processing

313

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Major components involved in ETL Processing


Design manager Lets developers define source-to-target mappings, transformations, process flows, and jobs Meta data management Provides a repository to define, document, and manage information about the ETL design and runtime processes Extract The process of reading data from a database. Transform The process of converting the extracted data Load The process of writing the data into the target database. Transport services ETL tools use network and file protocols to move data between source and target systems and in-memory protocols to move data between ETL run-time components. Administration and operation ETL utilities let administrators schedule, run, monitor ETL jobs, log all events, manage errors, recover from failures, reconcile outputs with source systems
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

314

ETL Tools
Provides facility to specify a large number of transformation rules with a GUI Generate programs to transform data Handle multiple data sources Handle data redundancy Generate metadata as output Most tools exploit parallelism by running on multiple low-cost servers in multi-threaded environment ETL Tools - Second-Generation PowerCentre/Mart from Informatica Data Mart Solution from Sagent Technology DataStage from Ascential
315

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Metadata Management

316

2009 Wipro Ltd - Confidential

What Is Metadata?
Metadata is Information...

That describes the WHAT, WHEN, WHO, WHERE, HOW of the data warehouse About the data being captured and loaded into the Warehouse Documented in IT tools that improves both business and technical understanding of data and data-related processes

317

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Importance Of Metadata
Locating Information Time spent in looking for information. How often information is found? What poor decisions were made based on the incomplete information?

How much money was lost or earned as a result? Interpreting information


How many times have businesses needed to rework or recall products? What impact does it have on the bottom line ? How many mistakes were due to misinterpretation of existing How much interpretation results form too much metadata? How much time is spent trying to determine if any of the metadata is accurate? Integrating information How various data perspectives connect together? How much time is spent trying to figure out that? How much does the inefficiency and lack of metadata affect decision making documentation?

318

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Requirements for DW Metadata Management


Provide a simple catalogue of business metadata descriptions and views Document/manage metadata descriptions from an integrated development environment Enable DW users to identify and invoke pre-built queries against the data stores Design and enhance new data models and schemas for the data warehouse Capture data transformation rules between the operational and data warehousing databases Provide change impact analysis, and update across these technologies
319
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Consumers of Metadata
Technical Users Warehouse administrator Application developer Business Users -Business metadata Meanings Definitions Business Rules Software Tools Used in DW life-cycle development Metadata requirements for each tool must be identified The tool-specific metadata should be analysed for inclusion in the enterprise metadata repository Previously captured metadata should be electronically transferred from the enterprise metadata repository to each individual tool

320

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Trends in the Metadata Management Tools


Third Party Bridging Tools Oracle Exchange
Technology of choice for a long list of repository, enterprise and workgroup vendors

Reischmann-Informatik-Toolbus
Features include facilitation of selective bridging of metadata

Ardent Software/ Dovetail Software -Interplay


Hub and Spoke solution for enabling metadata interoperability Ardent focussing on own engagements, not selling it as independent product

Informix's Metadata Plug-ins


Available with Ardent Datastage version 3.6.2 free of cost for Erwin, Oracle Designer, Sybase Powerdesigner, Brio, Microstrategy
321
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Trends in the Metadata Management Tools


Metadata Repositories IBM, Oracle and Microsoft to offer free or near-free basic repository services Enable organisations to reuse metadata across technologies Integrate DB design, data transformation and BI tools from different vendors Multi-tool vendors taking a bridged or federated rather than integrated approach to sharing metadata Both IBM and Oracle have multiple repositories for different lines of products e.g., One for AD and one for DW, with bridges between them

322

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Trends in the Metadata Management Tools


Metadata Interchange Standards CDIF (CASE Data Interchange Format)
Most frequently used interchange standard Addresses only a limited subset of metadata artifacts

OMG (Object Management Group)-CWM


XML-addresses context and data meaning, not presentation Can enable exchange over the web employing industry standards for storing and sharing programming data Will allow sharing of UML and MOF objects b/w various development tools and repositories

MDC (Metadata Coalition)


Based on XML/UML standards Promoted by Microsoft Along With 20 partners including Object Management Group (OMG), Oracle Carleton Group, CA-PLATINUM Technology (Founding Member), Viasoft
323
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

OLAP

324

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Agenda
OLAP Definition Distinction between OLTP and OLAP

MDDB Concepts
Implementation Techniques Architectures

Features
Representative Tools

6/19/2012

325

325

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

OLAP: On-Line Analytical Processing


OLAP can be defined as a technology which allows the users to view the aggregate data across measurements (like Maturity Amount, Interest Rate etc.) along with a set of related parameters called dimensions (like Product, Organization, Customer, etc.) Used interchangeably with BI Multidimensional view of data is the foundation of OLAP Users :Analysts, Decision makers
6/19/2012 326

326

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Distinction between OLTP and OLAP


OLTP System Source of data Operational data; OLTPs are the original source of the data To control and run fundamental business tasks A snapshot of ongoing business processes Short and fast inserts and updates initiated by end users
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

OLAP System Consolidation data; OLAP data comes from the various OLTP databases Decision support

Purpose of data

What the data reveals Inserts and Updates


6/19/2012

Multi-dimensional views of various kinds of business activities Periodic long-running batch jobs refresh the 327 data

327

MDDB Concepts
A multidimensional database is a computer software system designed to allow for efficient and convenient storage and retrieval of data that is intimately related and stored, viewed and analyzed from different perspectives (Dimensions). A hypercube represents a collection of multidimensional data. The edges of the cube are called dimensions Individual items within each dimensions are called members
328
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

RDBMS v/s MDDB: Increased Complexity...


Relational DBMS
MODEL MINI VAN MINI VAN MINI VAN MINI VAN MINI VAN MINI VAN MINI VAN MINI VAN MINI VAN SPORTS COUPE SPORTS COUPE SPORTS COUPE SPORTS COUPE SPORTS COUPE SPORTS COUPE SPORTS COUPE SPORTS COUPE SPORTS COUPE SEDAN SEDAN SEDAN ... COLOR BLUE BLUE BLUE RED RED RED WHITE WHITE WHITE BLUE BLUE BLUE RED RED RED WHITE WHITE WHITE BLUE BLUE BLUE DEALER Clyde Gleason Carr Clyde Gleason Carr Clyde Gleason Carr Clyde Gleason Carr Clyde Gleason Carr Clyde Gleason Carr Clyde Gleason Carr VOL. 6 3 2 5 3 1 3 1 4 3 3 3 4 3 6 2 3 5 4 3 2 ...

MDDB

Sales Volumes

M O D E L

Mini Van

Coupe Carr Gleason Clyde Blue Red White

Sedan

DEALERSHIP

COLOR

27 x 4 = 108 cells
329

3 x 3 x 3 = 27 cells

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Benefits of MDDB over RDBMS


Ease of Data Presentation & Navigation A great deal of information is gleaned immediately upon direct inspection of the array User is able to view data along presorted dimensions with data arranged in an inherently more organized, and accessible fashion than the one offered by the relational table. Storage Space Very low Space Consumption compared to Relational DB Performance Gives much better performance. Relational DB may give comparable results only through database tuning (indexing, keys etc), which may not be possible for ad-hoc queries. Ease of Maintenance No overhead as data is stored in the same way it is viewed. In Relational DB, indexes, sophisticated joins etc. are used which require considerable storage and maintenance
6/19/2012
330
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

330

Issues with MDDB

Sparsity
- Input data in applications are typically sparse -Increases with increased dimensions

Data Explosion
-Due to Sparsity -Due to Summarization

Performance
-Doesnt perform better than RDBMS at high data volumes (>20-30 GB)

6/19/2012
331
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

331

Issues with MDDB - Sparsity Example


If dimension members of different dimensions do not interact , then blank cell is left behind.
Employee Age
21 19 63 31 27 56 45 41 19
31 41 23 01 14 54 03 12 33

LAST NAME EMP# AGE SMITH 01 21 REGAN 12 Sales Volumes 19 FOX 31 63 Miini Van WELD 14 6 5 31 4 M O KELLY 54 3 5 27 Coupe D 5 E L LINK 03 56 4 3 2 Sedan KRANZ 41 45 Blue Red White LUCUS 33 COLOR 41 WEISS 23 19

Smith

Regan

Fox

L A S T N A M E

Weld

Kelly

Link

Kranz

Lucas

Weiss

EMPLOYEE #

6/19/2012
332
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

332

OLAP Features
Calculations applied across dimensions, through hierarchies and/or across members Trend analysis over sequential time periods, What-if scenarios. Slicing / Dicing subsets for on-screen viewing Rotation to new dimensional comparisons in the viewing area Drill-down/up along the hierarchy Reach-through / Drill-through to underlying detail data

6/19/2012
333
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

333

Features of OLAP - Rotation

Complex Queries & Sorts in Relational environment translated to simple rotation.


Sales Volumes

M O D E L

Mini Van

6 3 4
Blue

5 5 3
Red

4 5 2
White

Coupe

C O L O R ( ROTATE 90 )
o

Blue

6 5 4

3 5 5
MODEL

4 3 2
Sedan

Red

Sedan

White

Mini Van Coupe

COLOR

View #1

View #2

2 dimensional array has 2 views.


6/19/2012
334
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

334

Features of OLAP - Rotation


Sales Volumes

M O D E L

Mini Van Coupe Carr Gleason Clyde Blue Red White

Sedan

C O L O R

Blue

Red White Sedan Coupe Mini Van Carr Gleason Clyde

C O L O R

Blue

Red White Carr Gleason Clyde Mini Van Coupe Sedan

COLOR

( ROTATE 90 )

MODEL

( ROTATE 90 )

DEALERSHIP

( ROTATE 90 )

DEALERSHIP

DEALERSHIP

MODEL

View #1
D E A L E R S H I P D E A L E R S H I P

View #2

View #3

Carr Gleason Mini Van Coupe Sedan White Red Blue

Carr Gleason Blue Red White

Mini Van

Clyde

Clyde Mini Van Coupe Sedan

M O D E L

Coupe Blue Red White Clyde Gleason Carr

Sedan

COLOR

( ROTATE 90 )

MODEL

( ROTATE 90 )

DEALERSHIP

MODEL

COLOR

COLOR

View #4

View #5

View #6

3 dimensional array has 6 views.


6/19/2012
335
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

335

Features of OLAP - Slicing / Filtering


MDDB allows end user to quickly slice in on exact view of the data required.

Sales Volumes

M O D E L

Mini Van Mini Van

Coupe

Coupe Normal Metal Blue Blue

Carr Clyde

Carr Clyde

Normal Blue

Metal Blue

DEALERSHIP

COLOR
6/19/2012
336
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

336

Features of OLAP - Drill Down / Up

ORGANIZATION DIMENSION
REGION Midwest

DISTRICT

Chicago

St. Louis

Gary

DEALERSHIP

Clyde

Gleason

Carr

Levi

Lucas

Bolton

Sales at region/District/Dealership Level

Moving Up and moving down in a hierarchy is referred to as drill-up / roll-up and drill-down

6/19/2012
337
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

337

OLAP Reporting - Drill Down

Inflows ( Region , Year)


200 150 Inflows 100 ($M) 50 0 Year Year 1999 2000 Years

East West Central

6/19/2012
338
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

338

OLAP Reporting - Drill Down

Inflows ( Region , Year - Year 1999)


90 80 70 60 50 Inflows ( $M) 40 30 20 10 0

East West Central

1st Qtr

2nd Qtr 3rd Qtr Year 1999

4th Qtr

Drill-down from Year to Quarter


6/19/2012
339
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

339

OLAP Reporting - Drill Down

Inflows ( Region , Year - Year 1999 - 1st Qtr)


20 15 Inflows ( $M 10 ) 5

East West Central


January February March Year 1999

Drill-down from Quarter to Month

340

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Implementation Techniques -OLAP Architectures

MOLAP - Multidimensional OLAP


Multidimensional Databases for database and application logic layer

ROLAP - Relational OLAP


Access Data stored in relational Data Warehouse for OLAP Analysis. Database and Application logic provided as separate layers

HOLAP - Hybrid OLAP


OLAP Server routes queries first to MDDB, then to RDBMS and result processed on-the-fly in Server

DOLAP - Desk OLAP


Personal MDDB Server and application on the desktop

6/19/2012
341
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

341

MOLAP - MDDB storage

OLAP
Cube
OLAP Calculation Engine

Web Browser

OLAP Tools

OLAP Appli cations


6/19/2012
342
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

342

MOLAP - Features

Powerful analytical capabilities (e.g., financial, forecasting, statistical) Aggregation and calculation capabilities Read/write analytic applications Specialized data structures for
Maximum query performance. Optimum space utilization.
6/19/2012
343
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

343

ROLAP - Standard SQL storage

MDDB - Relational Mapping

Relational DW

Web Browser
OLAP Calculation Engine

SQL

OLAP Tools

OLAP Applications
6/19/2012
344
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

344

ROLAP - Features Three-tier hardware/software architecture:


GUI on client; multidimensional processing on midtier server; target database on database server Processing split between mid-tier & database servers

Ad hoc query capabilities to very large databases DW integration Data scalability

6/19/2012
345
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

345

HOLAP - Combination of RDBMS and MDDB


OLAP Cube
Any Client

Relational DW

Web Browser
OLAP Calculation Engine

SQL

OLAP Tools

OLAP Applications
6/19/2012
346
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

346

HOLAP - Features

RDBMS used for detailed data stored in large databases MDDB used for fast, read/write OLAP analysis and calculations Scalability of RDBMS and MDDB performance Calculation engine provides full analysis features Source of data transparent to end user

6/19/2012
347
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

347

Architecture Comparison

MOLAP
Definition

ROLAP

HOLAP
Hybrid OLAP = ROLAP + summary in MDDB Sparsity exists only in MDDB part To the necessary extent

MDDB OLAP = Relational OLAP = Transaction level data + Transaction level data + summary in MDDB summary in RDBMS Good Design 3 10 times High (May go beyond control. Estimation is very important) Fast - (Depends upon the size of the MDDB) No Sparsity To the necessary extent

Data explosion due to Sparsity Data explosion due to Summarization Query Execution Speed

Slow

Optimum - If the data is fetched from RDBMS then its like ROLAP otherwise like MOLAP. High: RDBMS + disk space + MDDB Server cost Large transactional data + frequent summary analysis

Cost

Medium: MDDB Server + large disk space cost

Low: Only RDBMS + disk space cost

Where to apply?

Small transactional Very large transactional data + complex model + data & it needs to be frequent summary viewed / sorted analysis

6/19/2012
348
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

348

Representative OLAP Tools:

Oracle Express Products Hyperion Essbase Cognos -PowerPlay Seagate - Holos SAS

Micro Strategy - DSS Agent Informix MetaCube Brio Query Business Objects / Web Intelligence

6/19/2012
349
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

349

Sample OLAP Applications

Sales Analysis Financial Analysis Profitability Analysis Performance Analysis Risk Management Profiling & Segmentation Scorecard Application NPA Management Strategic Planning Customer Relationship Management (CRM)
6/19/2012
350
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

350

Data Warehouse Testing

351

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Warehouse Testing Overview


There is an exponentially increasing cost associated with finding software defects later in the development lifecycle. In data warehousing, this is compounded because of the additional business costs of using incorrect data to make critical business decisions

The methodology required for testing a Data Warehouse is different from testing a typical transaction system

352

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Difference In Testing Data warehouse and Transaction System


Data warehouse testing is different on the following counts: User-Triggered vs. System triggered Volume of Test Data Possible scenarios/ Test Cases Programming for testing challenge

353

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Difference In Testing Data warehouse and Transaction System.


User-Triggered vs. System triggered

In data Warehouse, most of the testing is system triggered. Most of the production/Source system testing is the processing of individual transactions, which are driven by some input from the users (Application Form, Servicing Request.). There are very few test cycles, which cover the system-triggered scenarios (Like billing, Valuation.)

354

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Difference In Testing Data warehouse and Transaction System


Volume of Test Data The test data in a transaction system is a very small sample of the overall production data. Data Warehouse has typically large test data as one does try to fill-up maximum possible combination of dimensions and facts. Possible scenarios/ Test Cases In case of Data Warehouse, the permutations and combinations one can possibly test is virtually unlimited due to the core objective of Data Warehouse is to allow all possible views of data.

355

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Difference In Testing Data warehouse and Transaction System


Programming for testing challenge In case of transaction systems, users/business analysts typically test the output of the system. In case of data warehouse, most of the 'Data Warehouse data Quality testing' and ETL testing is done at backend by running separate stand-alone scripts. These scripts compare preTransformation to post Transformation of data.

356

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Warehouse Testing Process


Data-Warehouse testing is basically divided into two parts : 'Back-end' testing where the source systems data is compared to the endresult data in Loaded area 'Front-end' testing where the user checks the data by comparing their MIS with the data displayed by the end-user tools like OLAP. Testing phases consists of : Requirements testing Unit testing Integration testing Performance testing Acceptance testing

357

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Requirements testing
The main aim for doing Requirements testing is to check stated requirements for completeness. Requirements can be tested on following factors. Are the requirements Complete? Are the requirements Singular? Are the requirements Ambiguous? Are the requirements Developable? Are the requirements Testable?

358

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Unit Testing
Unit testing for data warehouses is WHITEBOX. It should check the ETL procedures/mappings/jobs and the reports developed. Unit testing the ETL procedures: Whether ETLs are accessing and picking up right data from right source.

All the data transformations are correct according to the business rules and data warehouse is correctly populated with the transformed data.
Testing the rejected records that dont fulfil transformation rules.

359

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Unit Testing
Unit Testing the Report data:

Verify Report data with source: Data present in a data warehouse will be stored at an aggregate level compare to source systems. QA team should verify the granular data stored in data warehouse against the source data available Field level data verification: QA team must understand the linkages for the fields displayed in the report and should trace back and compare that with the source systems Derivation formulae/calculation rules should be verified

360

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Integration Testing
Integration testing will involve following:

Sequence of ETLs jobs in batch. Initial loading of records on data warehouse. Incremental loading of records at a later date to verify the newly inserted or updated data. Testing the rejected records that dont fulfil transformation rules. Error log generation

361

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Performance Testing
Performance Testing should check for : ETL processes completing within time window.

Monitoring and measuring the data quality issues.


Refresh times for standard/complex reports.

362

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Acceptance testing
Here the system is tested with full functionality and is expected to function as in production. At the end of UAT, the system should be acceptable to the client for use in terms of ETL process integrity and business functionality and reporting.

363

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Questions

364

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Thank You

365

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Trends in the Metadata Management Tools


Metadata Interchange Standards CDIF (CASE Data Interchange Format)
Most frequently used interchange standard Addresses only a limited subset of metadata artifacts

OMG (Object Management Group)-CWM


XML-addresses context and data meaning, not presentation Can enable exchange over the web employing industry standards for storing and sharing programming data Will allow sharing of UML and MOF objects b/w various development tools and repositories

MDC (Metadata Coalition)


Based on XML/UML standards Promoted by Microsoft Along With 20 partners including Object Management Group (OMG), Oracle Carleton Group, CA-PLATINUM Technology (Founding Member), Viasoft
366
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

OLAP

367

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Agenda
OLAP Definition Distinction between OLTP and OLAP

MDDB Concepts
Implementation Techniques Architectures

Features
Representative Tools

6/19/2012

368

368

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

OLAP: On-Line Analytical Processing


OLAP can be defined as a technology which allows the users to view the aggregate data across measurements (like Maturity Amount, Interest Rate etc.) along with a set of related parameters called dimensions (like Product, Organization, Customer, etc.) Used interchangeably with BI Multidimensional view of data is the foundation of OLAP Users :Analysts, Decision makers
6/19/2012 369

369

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Distinction between OLTP and OLAP


OLTP System Source of data Operational data; OLTPs are the original source of the data To control and run fundamental business tasks A snapshot of ongoing business processes Short and fast inserts and updates initiated by end users
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

OLAP System Consolidation data; OLAP data comes from the various OLTP databases Decision support

Purpose of data

What the data reveals Inserts and Updates


6/19/2012

Multi-dimensional views of various kinds of business activities Periodic long-running batch jobs refresh the 370 data

370

MDDB Concepts
A multidimensional database is a computer software system designed to allow for efficient and convenient storage and retrieval of data that is intimately related and stored, viewed and analyzed from different perspectives (Dimensions). A hypercube represents a collection of multidimensional data. The edges of the cube are called dimensions Individual items within each dimensions are called members
371
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

RDBMS v/s MDDB: Increased Complexity...


Relational DBMS
MODEL MINI VAN MINI VAN MINI VAN MINI VAN MINI VAN MINI VAN MINI VAN MINI VAN MINI VAN SPORTS COUPE SPORTS COUPE SPORTS COUPE SPORTS COUPE SPORTS COUPE SPORTS COUPE SPORTS COUPE SPORTS COUPE SPORTS COUPE SEDAN SEDAN SEDAN ... COLOR BLUE BLUE BLUE RED RED RED WHITE WHITE WHITE BLUE BLUE BLUE RED RED RED WHITE WHITE WHITE BLUE BLUE BLUE DEALER Clyde Gleason Carr Clyde Gleason Carr Clyde Gleason Carr Clyde Gleason Carr Clyde Gleason Carr Clyde Gleason Carr Clyde Gleason Carr VOL. 6 3 2 5 3 1 3 1 4 3 3 3 4 3 6 2 3 5 4 3 2 ...

MDDB

Sales Volumes

M O D E L

Mini Van

Coupe Carr Gleason Clyde Blue Red White

Sedan

DEALERSHIP

COLOR

27 x 4 = 108 cells
372

3 x 3 x 3 = 27 cells

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Benefits of MDDB over RDBMS


Ease of Data Presentation & Navigation A great deal of information is gleaned immediately upon direct inspection of the array User is able to view data along presorted dimensions with data arranged in an inherently more organized, and accessible fashion than the one offered by the relational table. Storage Space Very low Space Consumption compared to Relational DB Performance Gives much better performance. Relational DB may give comparable results only through database tuning (indexing, keys etc), which may not be possible for ad-hoc queries. Ease of Maintenance No overhead as data is stored in the same way it is viewed. In Relational DB, indexes, sophisticated joins etc. are used which require considerable storage and maintenance
6/19/2012
373
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

373

Issues with MDDB

Sparsity
- Input data in applications are typically sparse -Increases with increased dimensions

Data Explosion
-Due to Sparsity -Due to Summarization

Performance
-Doesnt perform better than RDBMS at high data volumes (>20-30 GB)

6/19/2012
374
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

374

Issues with MDDB - Sparsity Example


If dimension members of different dimensions do not interact , then blank cell is left behind.
Employee Age
21 19 63 31 27 56 45 41 19
31 41 23 01 14 54 03 12 33

LAST NAME EMP# AGE SMITH 01 21 REGAN 12 Sales Volumes 19 FOX 31 63 Miini Van WELD 14 6 5 31 4 M O KELLY 54 3 5 27 Coupe D 5 E L LINK 03 56 4 3 2 Sedan KRANZ 41 45 Blue Red White LUCUS 33 COLOR 41 WEISS 23 19

Smith

Regan

Fox

L A S T N A M E

Weld

Kelly

Link

Kranz

Lucas

Weiss

EMPLOYEE #

6/19/2012
375
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

375

OLAP Features
Calculations applied across dimensions, through hierarchies and/or across members Trend analysis over sequential time periods, What-if scenarios. Slicing / Dicing subsets for on-screen viewing Rotation to new dimensional comparisons in the viewing area Drill-down/up along the hierarchy Reach-through / Drill-through to underlying detail data

6/19/2012
376
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

376

Features of OLAP - Rotation

Complex Queries & Sorts in Relational environment translated to simple rotation.


Sales Volumes

M O D E L

Mini Van

6 3 4
Blue

5 5 3
Red

4 5 2
White

Coupe

C O L O R ( ROTATE 90 )
o

Blue

6 5 4

3 5 5
MODEL

4 3 2
Sedan

Red

Sedan

White

Mini Van Coupe

COLOR

View #1

View #2

2 dimensional array has 2 views.


6/19/2012
377
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

377

Features of OLAP - Rotation


Sales Volumes

M O D E L

Mini Van Coupe Carr Gleason Clyde Blue Red White

Sedan

C O L O R

Blue

Red White Sedan Coupe Mini Van Carr Gleason Clyde

C O L O R

Blue

Red White Carr Gleason Clyde Mini Van Coupe Sedan

COLOR

( ROTATE 90 )

MODEL

( ROTATE 90 )

DEALERSHIP

( ROTATE 90 )

DEALERSHIP

DEALERSHIP

MODEL

View #1
D E A L E R S H I P D E A L E R S H I P

View #2

View #3

Carr Gleason Mini Van Coupe Sedan White Red Blue

Carr Gleason Blue Red White

Mini Van

Clyde

Clyde Mini Van Coupe Sedan

M O D E L

Coupe Blue Red White Clyde Gleason Carr

Sedan

COLOR

( ROTATE 90 )

MODEL

( ROTATE 90 )

DEALERSHIP

MODEL

COLOR

COLOR

View #4

View #5

View #6

3 dimensional array has 6 views.


6/19/2012
378
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

378

Features of OLAP - Slicing / Filtering


MDDB allows end user to quickly slice in on exact view of the data required.

Sales Volumes

M O D E L

Mini Van Mini Van

Coupe

Coupe Normal Metal Blue Blue

Carr Clyde

Carr Clyde

Normal Blue

Metal Blue

DEALERSHIP

COLOR
6/19/2012
379
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

379

Features of OLAP - Drill Down / Up

ORGANIZATION DIMENSION
REGION Midwest

DISTRICT

Chicago

St. Louis

Gary

DEALERSHIP

Clyde

Gleason

Carr

Levi

Lucas

Bolton

Sales at region/District/Dealership Level

Moving Up and moving down in a hierarchy is referred to as drill-up / roll-up and drill-down

6/19/2012
380
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

380

OLAP Reporting - Drill Down

Inflows ( Region , Year)


200 150 Inflows 100 ($M) 50 0 Year Year 1999 2000 Years

East West Central

6/19/2012
381
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

381

OLAP Reporting - Drill Down

Inflows ( Region , Year - Year 1999)


90 80 70 60 50 Inflows ( $M) 40 30 20 10 0

East West Central

1st Qtr

2nd Qtr 3rd Qtr Year 1999

4th Qtr

Drill-down from Year to Quarter


6/19/2012
382
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

382

OLAP Reporting - Drill Down

Inflows ( Region , Year - Year 1999 - 1st Qtr)


20 15 Inflows ( $M 10 ) 5

East West Central


January February March Year 1999

Drill-down from Quarter to Month

383

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Implementation Techniques -OLAP Architectures

MOLAP - Multidimensional OLAP


Multidimensional Databases for database and application logic layer

ROLAP - Relational OLAP


Access Data stored in relational Data Warehouse for OLAP Analysis. Database and Application logic provided as separate layers

HOLAP - Hybrid OLAP


OLAP Server routes queries first to MDDB, then to RDBMS and result processed on-the-fly in Server

DOLAP - Desk OLAP


Personal MDDB Server and application on the desktop

6/19/2012
384
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

384

MOLAP - MDDB storage

OLAP
Cube
OLAP Calculation Engine

Web Browser

OLAP Tools

OLAP Appli cations


6/19/2012
385
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

385

MOLAP - Features

Powerful analytical capabilities (e.g., financial, forecasting, statistical) Aggregation and calculation capabilities Read/write analytic applications Specialized data structures for
Maximum query performance. Optimum space utilization.
6/19/2012
386
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

386

ROLAP - Standard SQL storage

MDDB - Relational Mapping

Relational DW

Web Browser
OLAP Calculation Engine

SQL

OLAP Tools

OLAP Applications
6/19/2012
387
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

387

ROLAP - Features Three-tier hardware/software architecture:


GUI on client; multidimensional processing on midtier server; target database on database server Processing split between mid-tier & database servers

Ad hoc query capabilities to very large databases DW integration Data scalability

6/19/2012
388
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

388

HOLAP - Combination of RDBMS and MDDB


OLAP Cube
Any Client

Relational DW

Web Browser
OLAP Calculation Engine

SQL

OLAP Tools

OLAP Applications
6/19/2012
389
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

389

HOLAP - Features

RDBMS used for detailed data stored in large databases MDDB used for fast, read/write OLAP analysis and calculations Scalability of RDBMS and MDDB performance Calculation engine provides full analysis features Source of data transparent to end user

6/19/2012
390
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

390

Architecture Comparison

MOLAP
Definition

ROLAP

HOLAP
Hybrid OLAP = ROLAP + summary in MDDB Sparsity exists only in MDDB part To the necessary extent

MDDB OLAP = Relational OLAP = Transaction level data + Transaction level data + summary in MDDB summary in RDBMS Good Design 3 10 times High (May go beyond control. Estimation is very important) Fast - (Depends upon the size of the MDDB) No Sparsity To the necessary extent

Data explosion due to Sparsity Data explosion due to Summarization Query Execution Speed

Slow

Optimum - If the data is fetched from RDBMS then its like ROLAP otherwise like MOLAP. High: RDBMS + disk space + MDDB Server cost Large transactional data + frequent summary analysis

Cost

Medium: MDDB Server + large disk space cost

Low: Only RDBMS + disk space cost

Where to apply?

Small transactional Very large transactional data + complex model + data & it needs to be frequent summary viewed / sorted analysis

6/19/2012
391
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

391

Representative OLAP Tools:

Oracle Express Products Hyperion Essbase Cognos -PowerPlay Seagate - Holos SAS

Micro Strategy - DSS Agent Informix MetaCube Brio Query Business Objects / Web Intelligence

6/19/2012
392
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

392

Sample OLAP Applications

Sales Analysis Financial Analysis Profitability Analysis Performance Analysis Risk Management Profiling & Segmentation Scorecard Application NPA Management Strategic Planning Customer Relationship Management (CRM)
6/19/2012
393
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

393

Data Warehouse Testing

394

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Warehouse Testing Overview


There is an exponentially increasing cost associated with finding software defects later in the development lifecycle. In data warehousing, this is compounded because of the additional business costs of using incorrect data to make critical business decisions

The methodology required for testing a Data Warehouse is different from testing a typical transaction system

395

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Difference In Testing Data warehouse and Transaction System


Data warehouse testing is different on the following counts: User-Triggered vs. System triggered Volume of Test Data Possible scenarios/ Test Cases Programming for testing challenge

396

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Difference In Testing Data warehouse and Transaction System.


User-Triggered vs. System triggered

In data Warehouse, most of the testing is system triggered. Most of the production/Source system testing is the processing of individual transactions, which are driven by some input from the users (Application Form, Servicing Request.). There are very few test cycles, which cover the system-triggered scenarios (Like billing, Valuation.)

397

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Difference In Testing Data warehouse and Transaction System


Volume of Test Data The test data in a transaction system is a very small sample of the overall production data. Data Warehouse has typically large test data as one does try to fill-up maximum possible combination of dimensions and facts. Possible scenarios/ Test Cases In case of Data Warehouse, the permutations and combinations one can possibly test is virtually unlimited due to the core objective of Data Warehouse is to allow all possible views of data.

398

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Difference In Testing Data warehouse and Transaction System


Programming for testing challenge In case of transaction systems, users/business analysts typically test the output of the system. In case of data warehouse, most of the 'Data Warehouse data Quality testing' and ETL testing is done at backend by running separate stand-alone scripts. These scripts compare preTransformation to post Transformation of data.

399

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Warehouse Testing Process


Data-Warehouse testing is basically divided into two parts : 'Back-end' testing where the source systems data is compared to the endresult data in Loaded area 'Front-end' testing where the user checks the data by comparing their MIS with the data displayed by the end-user tools like OLAP. Testing phases consists of : Requirements testing Unit testing Integration testing Performance testing Acceptance testing

400

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Requirements testing
The main aim for doing Requirements testing is to check stated requirements for completeness. Requirements can be tested on following factors. Are the requirements Complete? Are the requirements Singular? Are the requirements Ambiguous? Are the requirements Developable? Are the requirements Testable?

401

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Unit Testing
Unit testing for data warehouses is WHITEBOX. It should check the ETL procedures/mappings/jobs and the reports developed. Unit testing the ETL procedures: Whether ETLs are accessing and picking up right data from right source.

All the data transformations are correct according to the business rules and data warehouse is correctly populated with the transformed data.
Testing the rejected records that dont fulfil transformation rules.

402

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Unit Testing
Unit Testing the Report data:

Verify Report data with source: Data present in a data warehouse will be stored at an aggregate level compare to source systems. QA team should verify the granular data stored in data warehouse against the source data available Field level data verification: QA team must understand the linkages for the fields displayed in the report and should trace back and compare that with the source systems Derivation formulae/calculation rules should be verified

403

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Integration Testing
Integration testing will involve following:

Sequence of ETLs jobs in batch. Initial loading of records on data warehouse. Incremental loading of records at a later date to verify the newly inserted or updated data. Testing the rejected records that dont fulfil transformation rules. Error log generation

404

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Performance Testing
Performance Testing should check for : ETL processes completing within time window.

Monitoring and measuring the data quality issues.


Refresh times for standard/complex reports.

405

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Acceptance testing
Here the system is tested with full functionality and is expected to function as in production. At the end of UAT, the system should be acceptable to the client for use in terms of ETL process integrity and business functionality and reporting.

406

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Questions

407

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Thank You

408

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Components of Warehouse
Source Tables: These are real-time, volatile data in relational databases for transaction processing (OLTP). These can be any relational databases or flat files. ETL Tools: To extract, cleansing, transform (aggregates, joins) and load the data from sources to target. Maintenance and Administration Tools: To authorize and monitor access to the data, set-up users. Scheduling jobs to run on offshore periods. Modeling Tools: Used for data warehouse design for high-performance using dimensional data modeling technique, mapping the source and target files. Databases: Target databases and data marts, which are part of data warehouse. These are structured for analysis and reporting purposes. End-user tools for analysis and reporting: get the reports and analyze the data from target tables. Different types of Querying, Data Mining, OLAP tools are used for this purpose.

409

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Warehouse Architecture


This is a basic design, where there are source files, which are loaded to a warehouse and users query the data for different purposes.

This has a staging area, where the data after cleansing, transforming is loaded and tested here. Later is directly loaded to the target database/warehouse. Which is divided to data marts and can be accessed by different users for their reporting and analyzing purposes.

410

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Modeling
Effective way of using a Data Warehouse

411

2009 Wipro Ltd - Confidential

Data Modeling
Commonly E-R Data Model is used in OLTP, In OLAP Dimensional Data Model is used commonly. E-R (Entity-Relationship) Data Model
Entity: Object that can be observed and classified based on its properties and characteristics. Like employee, book, student Relationship: relating entities to other entities.

Different Perceptive of Data Modeling.


o Conceptual Data Model o Logical Data Model o Physical Data Model

Types of Dimensional Data Models most commonly used:


o Star Schema o Snowflake Schema
412
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Terms used in Dimensional Data Model


To understand dimensional data modeling, let's define some of the terms commonly used in this type of modeling: Dimension: A category of information. For example, the time dimension. Attribute: A unique level within a dimension. For example, Month is an attribute in the Time Dimension. Hierarchy: The specification of levels that represents relationship between different attributes within a dimension. For example, one possible hierarchy in the Time dimension is Year Quarter Month Day. Fact Table: A table that contains the measures of interest. Lookup Table: It provides the detailed information about the attributes. For example, the lookup table for the Quarter attribute would include a list of all of the quarters available in the data warehouse. Surrogate Keys: To avoid the data integrity, surrogate keys are used. They are helpful for Slow Changing Dimensions and act as index/primary keys.
A dimensional model includes fact tables and lookup tables. Fact tables connect to one or more lookup tables, but fact tables do not have direct relationships to one another. Dimensions and hierarchies are represented by lookup tables. Attributes are the non-key columns in the lookup tables.
413
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Star Schema
Dimension Table
product prodId p1 p2 name price bolt 10 nut 5

Dimension Table
store storeId c1 c2 c3 city nyc sfo la

Fact Table
sale oderId date o100 1/7/97 o102 2/7/97 105 3/8/97 custId 53 53 111 prodId p1 p2 p1 storeId c1 c1 c3 qty 1 2 5 amt 12 11 50

Dimension Table
customer custId 53 81 111 name joe fred sally address 10 main 12 main 80 willow city sfo sfo la

414

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Snowflake Schema
Dimension Table Fact Table
store storeId s5 s7 s9 cityId sfo sfo la tId t1 t2 t1 mgr joe fred nancy
sType tId t1 t2 city size small large location downtown suburbs regId north south

Dimension Table
cityId pop sfo 1M la 5M

The star and snowflake schema are most commonly found in dimensional data warehouses and data marts where speed of data retrieval is more important than the efficiency of data manipulations. As such, the tables in these schema are not normalized much, and are frequently designed at a level of normalization short of third normal form.

region regId name north cold region south warm region

415

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Overview of Data Cleansing

416

2009 Wipro Ltd - Confidential

The Need For Data Quality


Difficulty in decision making Time delays in operation Organizational mistrust Data ownership conflicts Customer attrition Costs associated with error detection error rework customer service fixing customer problems

417

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Six Steps To Data Quality


Understand Information Flow In Organization
Identify authoritative data sources Interview Employees & Customers

Identify Potential Problem Areas & Asses Impact

Data Entry Points


Cost of bad data

Measure Quality Of Data

Use business rule discovery tools to identify data with

inconsistent, missing, incomplete, duplicate or incorrect values


Use data cleansing tools to clean data at the source Load only clean data into the data warehouse

Clean & Load Data

Continuous Monitoring

Schedule Periodic Cleansing of Source Data

Identify Areas of Improvement

Identify & Correct Cause of Defects Refine data capture mechanisms at source Educate users on importance of DQ
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

418

Data Quality Solution


Customized Programs Strengths: Addresses specific needs No bulky one time investment Limitations Tons of Custom programs in different environments are difficult to manage Minor alterations demand coding efforts Data Quality Assessment tools Strength Provide automated assessment Limitation No measure of data accuracy

419

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Quality Solution


Business Rule Discovery tools Strengths Detect Correlation in data values Can detect Patterns of behavior that indicate fraud Limitations Not all variables can be discovered Some discovered rules might not be pertinent There may be performance problems with large files or with many fields. Data Reengineering & Cleansing tools Strengths Usually are integrated packages with cleansing features as Addon Limitations Error prevention at source is usually absent The ETL tools have limited cleansing facilities
420
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Tools In The Market


Business Rule Discovery Tools Integrity Data Reengineering Tool from Vality Technology Trillium Software System from Harte -Hanks Data Technologies Migration Architect from DB Star Data Reengineering & Cleansing Tools Carlton Pureview from Oracle ETI-Extract from Evolutionary Technologies PowerMart from Informatica Corp Sagent Data Mart from Sagent Technology Data Quality Assessment Tools Migration Architect, Evoke Axio from Evoke Software Wizrule from Wizsoft Name & Address Cleansing Tools Centrus Suite from Sagent I.d.centric from First Logic

421

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Extraction, Transformation, Load

422

2009 Wipro Ltd - Confidential

ETL Architecture

Visitors

Web Browsers

The Internet

External Data Demographics, Household, Webographics, Income

Staging Area
Web Server Logs & E-comm Transaction Data Flat Files

Meta Data Repository

Scheduled Extraction

RDBMS

Clean Transform Match Merge

Scheduled Loading

Enterprise Data Warehouse

Other OLTP Systems

Data Collection

Data Extraction

Data Transformation

Data Loading

Data Storage & Integration

423

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

ETL Architecture
Data Extraction:
Rummages through a file or database Uses some criteria for selection Identifies qualified data and Transports the data over onto another file or database

Data transformation
Integrating dissimilar data types Changing codes Adding a time attribute Summarizing data Calculating derived values Renormalizing data

Data Extraction Cleanup


Restructuring of records or fields Removal of Operational-only data Supply of missing field values Data Integrity checks Data Consistency and Range checks, etc...

Data loading
Initial and incremental loading Updation of metadata

424

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Why ETL ?
Companies have valuable data lying around throughout their networks that needs to be moved from one place to another. The data lies in all sorts of heterogeneous systems,and therefore in all sorts of formats.

To solve the problem, companies use extract, transform and load (ETL) software.
The data used in ETL processes can come from any source: a mainframe application, an ERP application, a CRM tool, a flat file, and an Excel spreadsheet.

425

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Major components involved in ETL Processing

426

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Major components involved in ETL Processing


Design manager Lets developers define source-to-target mappings, transformations, process flows, and jobs Meta data management Provides a repository to define, document, and manage information about the ETL design and runtime processes Extract The process of reading data from a database. Transform The process of converting the extracted data Load The process of writing the data into the target database. Transport services ETL tools use network and file protocols to move data between source and target systems and in-memory protocols to move data between ETL run-time components. Administration and operation ETL utilities let administrators schedule, run, monitor ETL jobs, log all events, manage errors, recover from failures, reconcile outputs with source systems
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

427

ETL Tools
Provides facility to specify a large number of transformation rules with a GUI Generate programs to transform data Handle multiple data sources Handle data redundancy Generate metadata as output Most tools exploit parallelism by running on multiple low-cost servers in multi-threaded environment ETL Tools - Second-Generation PowerCentre/Mart from Informatica Data Mart Solution from Sagent Technology DataStage from Ascential
428

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Metadata Management

429

2009 Wipro Ltd - Confidential

What Is Metadata?
Metadata is Information...

That describes the WHAT, WHEN, WHO, WHERE, HOW of the data warehouse About the data being captured and loaded into the Warehouse Documented in IT tools that improves both business and technical understanding of data and data-related processes

430

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Importance Of Metadata
Locating Information Time spent in looking for information. How often information is found? What poor decisions were made based on the incomplete information?

How much money was lost or earned as a result? Interpreting information


How many times have businesses needed to rework or recall products? What impact does it have on the bottom line ? How many mistakes were due to misinterpretation of existing How much interpretation results form too much metadata? How much time is spent trying to determine if any of the metadata is accurate? Integrating information How various data perspectives connect together? How much time is spent trying to figure out that? How much does the inefficiency and lack of metadata affect decision making documentation?

431

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Requirements for DW Metadata Management


Provide a simple catalogue of business metadata descriptions and views Document/manage metadata descriptions from an integrated development environment Enable DW users to identify and invoke pre-built queries against the data stores Design and enhance new data models and schemas for the data warehouse Capture data transformation rules between the operational and data warehousing databases Provide change impact analysis, and update across these technologies
432
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Consumers of Metadata
Technical Users Warehouse administrator Application developer Business Users -Business metadata Meanings Definitions Business Rules Software Tools Used in DW life-cycle development Metadata requirements for each tool must be identified The tool-specific metadata should be analysed for inclusion in the enterprise metadata repository Previously captured metadata should be electronically transferred from the enterprise metadata repository to each individual tool

433

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Trends in the Metadata Management Tools


Third Party Bridging Tools Oracle Exchange
Technology of choice for a long list of repository, enterprise and workgroup vendors

Reischmann-Informatik-Toolbus
Features include facilitation of selective bridging of metadata

Ardent Software/ Dovetail Software -Interplay


Hub and Spoke solution for enabling metadata interoperability Ardent focussing on own engagements, not selling it as independent product

Informix's Metadata Plug-ins


Available with Ardent Datastage version 3.6.2 free of cost for Erwin, Oracle Designer, Sybase Powerdesigner, Brio, Microstrategy
434
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Trends in the Metadata Management Tools


Metadata Repositories IBM, Oracle and Microsoft to offer free or near-free basic repository services Enable organisations to reuse metadata across technologies Integrate DB design, data transformation and BI tools from different vendors Multi-tool vendors taking a bridged or federated rather than integrated approach to sharing metadata Both IBM and Oracle have multiple repositories for different lines of products e.g., One for AD and one for DW, with bridges between them

435

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Trends in the Metadata Management Tools


Metadata Interchange Standards CDIF (CASE Data Interchange Format)
Most frequently used interchange standard Addresses only a limited subset of metadata artifacts

OMG (Object Management Group)-CWM


XML-addresses context and data meaning, not presentation Can enable exchange over the web employing industry standards for storing and sharing programming data Will allow sharing of UML and MOF objects b/w various development tools and repositories

MDC (Metadata Coalition)


Based on XML/UML standards Promoted by Microsoft Along With 20 partners including Object Management Group (OMG), Oracle Carleton Group, CA-PLATINUM Technology (Founding Member), Viasoft
436
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

OLAP

437

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Agenda
OLAP Definition Distinction between OLTP and OLAP

MDDB Concepts
Implementation Techniques Architectures

Features
Representative Tools

6/19/2012

438

438

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

OLAP: On-Line Analytical Processing


OLAP can be defined as a technology which allows the users to view the aggregate data across measurements (like Maturity Amount, Interest Rate etc.) along with a set of related parameters called dimensions (like Product, Organization, Customer, etc.) Used interchangeably with BI Multidimensional view of data is the foundation of OLAP Users :Analysts, Decision makers
6/19/2012 439

439

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Distinction between OLTP and OLAP


OLTP System Source of data Operational data; OLTPs are the original source of the data To control and run fundamental business tasks A snapshot of ongoing business processes Short and fast inserts and updates initiated by end users
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

OLAP System Consolidation data; OLAP data comes from the various OLTP databases Decision support

Purpose of data

What the data reveals Inserts and Updates


6/19/2012

Multi-dimensional views of various kinds of business activities Periodic long-running batch jobs refresh the 440 data

440

MDDB Concepts
A multidimensional database is a computer software system designed to allow for efficient and convenient storage and retrieval of data that is intimately related and stored, viewed and analyzed from different perspectives (Dimensions). A hypercube represents a collection of multidimensional data. The edges of the cube are called dimensions Individual items within each dimensions are called members
441
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

RDBMS v/s MDDB: Increased Complexity...


Relational DBMS
MODEL MINI VAN MINI VAN MINI VAN MINI VAN MINI VAN MINI VAN MINI VAN MINI VAN MINI VAN SPORTS COUPE SPORTS COUPE SPORTS COUPE SPORTS COUPE SPORTS COUPE SPORTS COUPE SPORTS COUPE SPORTS COUPE SPORTS COUPE SEDAN SEDAN SEDAN ... COLOR BLUE BLUE BLUE RED RED RED WHITE WHITE WHITE BLUE BLUE BLUE RED RED RED WHITE WHITE WHITE BLUE BLUE BLUE DEALER Clyde Gleason Carr Clyde Gleason Carr Clyde Gleason Carr Clyde Gleason Carr Clyde Gleason Carr Clyde Gleason Carr Clyde Gleason Carr VOL. 6 3 2 5 3 1 3 1 4 3 3 3 4 3 6 2 3 5 4 3 2 ...

MDDB

Sales Volumes

M O D E L

Mini Van

Coupe Carr Gleason Clyde Blue Red White

Sedan

DEALERSHIP

COLOR

27 x 4 = 108 cells
442

3 x 3 x 3 = 27 cells

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Benefits of MDDB over RDBMS


Ease of Data Presentation & Navigation A great deal of information is gleaned immediately upon direct inspection of the array User is able to view data along presorted dimensions with data arranged in an inherently more organized, and accessible fashion than the one offered by the relational table. Storage Space Very low Space Consumption compared to Relational DB Performance Gives much better performance. Relational DB may give comparable results only through database tuning (indexing, keys etc), which may not be possible for ad-hoc queries. Ease of Maintenance No overhead as data is stored in the same way it is viewed. In Relational DB, indexes, sophisticated joins etc. are used which require considerable storage and maintenance
6/19/2012
443
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

443

Issues with MDDB

Sparsity
- Input data in applications are typically sparse -Increases with increased dimensions

Data Explosion
-Due to Sparsity -Due to Summarization

Performance
-Doesnt perform better than RDBMS at high data volumes (>20-30 GB)

6/19/2012
444
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

444

Issues with MDDB - Sparsity Example


If dimension members of different dimensions do not interact , then blank cell is left behind.
Employee Age
21 19 63 31 27 56 45 41 19
31 41 23 01 14 54 03 12 33

LAST NAME EMP# AGE SMITH 01 21 REGAN 12 Sales Volumes 19 FOX 31 63 Miini Van WELD 14 6 5 31 4 M O KELLY 54 3 5 27 Coupe D 5 E L LINK 03 56 4 3 2 Sedan KRANZ 41 45 Blue Red White LUCUS 33 COLOR 41 WEISS 23 19

Smith

Regan

Fox

L A S T N A M E

Weld

Kelly

Link

Kranz

Lucas

Weiss

EMPLOYEE #

6/19/2012
445
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

445

OLAP Features
Calculations applied across dimensions, through hierarchies and/or across members Trend analysis over sequential time periods, What-if scenarios. Slicing / Dicing subsets for on-screen viewing Rotation to new dimensional comparisons in the viewing area Drill-down/up along the hierarchy Reach-through / Drill-through to underlying detail data

6/19/2012
446
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

446

Features of OLAP - Rotation

Complex Queries & Sorts in Relational environment translated to simple rotation.


Sales Volumes

M O D E L

Mini Van

6 3 4
Blue

5 5 3
Red

4 5 2
White

Coupe

C O L O R ( ROTATE 90 )
o

Blue

6 5 4

3 5 5
MODEL

4 3 2
Sedan

Red

Sedan

White

Mini Van Coupe

COLOR

View #1

View #2

2 dimensional array has 2 views.


6/19/2012
447
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

447

Features of OLAP - Rotation


Sales Volumes

M O D E L

Mini Van Coupe Carr Gleason Clyde Blue Red White

Sedan

C O L O R

Blue

Red White Sedan Coupe Mini Van Carr Gleason Clyde

C O L O R

Blue

Red White Carr Gleason Clyde Mini Van Coupe Sedan

COLOR

( ROTATE 90 )

MODEL

( ROTATE 90 )

DEALERSHIP

( ROTATE 90 )

DEALERSHIP

DEALERSHIP

MODEL

View #1
D E A L E R S H I P D E A L E R S H I P

View #2

View #3

Carr Gleason Mini Van Coupe Sedan White Red Blue

Carr Gleason Blue Red White

Mini Van

Clyde

Clyde Mini Van Coupe Sedan

M O D E L

Coupe Blue Red White Clyde Gleason Carr

Sedan

COLOR

( ROTATE 90 )

MODEL

( ROTATE 90 )

DEALERSHIP

MODEL

COLOR

COLOR

View #4

View #5

View #6

3 dimensional array has 6 views.


6/19/2012
448
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

448

Features of OLAP - Slicing / Filtering


MDDB allows end user to quickly slice in on exact view of the data required.

Sales Volumes

M O D E L

Mini Van Mini Van

Coupe

Coupe Normal Metal Blue Blue

Carr Clyde

Carr Clyde

Normal Blue

Metal Blue

DEALERSHIP

COLOR
6/19/2012
449
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

449

Features of OLAP - Drill Down / Up

ORGANIZATION DIMENSION
REGION Midwest

DISTRICT

Chicago

St. Louis

Gary

DEALERSHIP

Clyde

Gleason

Carr

Levi

Lucas

Bolton

Sales at region/District/Dealership Level

Moving Up and moving down in a hierarchy is referred to as drill-up / roll-up and drill-down

6/19/2012
450
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

450

OLAP Reporting - Drill Down

Inflows ( Region , Year)


200 150 Inflows 100 ($M) 50 0 Year Year 1999 2000 Years

East West Central

6/19/2012
451
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

451

OLAP Reporting - Drill Down

Inflows ( Region , Year - Year 1999)


90 80 70 60 50 Inflows ( $M) 40 30 20 10 0

East West Central

1st Qtr

2nd Qtr 3rd Qtr Year 1999

4th Qtr

Drill-down from Year to Quarter


6/19/2012
452
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

452

OLAP Reporting - Drill Down

Inflows ( Region , Year - Year 1999 - 1st Qtr)


20 15 Inflows ( $M 10 ) 5

East West Central


January February March Year 1999

Drill-down from Quarter to Month

453

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Implementation Techniques -OLAP Architectures

MOLAP - Multidimensional OLAP


Multidimensional Databases for database and application logic layer

ROLAP - Relational OLAP


Access Data stored in relational Data Warehouse for OLAP Analysis. Database and Application logic provided as separate layers

HOLAP - Hybrid OLAP


OLAP Server routes queries first to MDDB, then to RDBMS and result processed on-the-fly in Server

DOLAP - Desk OLAP


Personal MDDB Server and application on the desktop

6/19/2012
454
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

454

MOLAP - MDDB storage

OLAP
Cube
OLAP Calculation Engine

Web Browser

OLAP Tools

OLAP Appli cations


6/19/2012
455
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

455

MOLAP - Features

Powerful analytical capabilities (e.g., financial, forecasting, statistical) Aggregation and calculation capabilities Read/write analytic applications Specialized data structures for
Maximum query performance. Optimum space utilization.
6/19/2012
456
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

456

ROLAP - Standard SQL storage

MDDB - Relational Mapping

Relational DW

Web Browser
OLAP Calculation Engine

SQL

OLAP Tools

OLAP Applications
6/19/2012
457
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

457

ROLAP - Features Three-tier hardware/software architecture:


GUI on client; multidimensional processing on midtier server; target database on database server Processing split between mid-tier & database servers

Ad hoc query capabilities to very large databases DW integration Data scalability

6/19/2012
458
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

458

HOLAP - Combination of RDBMS and MDDB


OLAP Cube
Any Client

Relational DW

Web Browser
OLAP Calculation Engine

SQL

OLAP Tools

OLAP Applications
6/19/2012
459
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

459

HOLAP - Features

RDBMS used for detailed data stored in large databases MDDB used for fast, read/write OLAP analysis and calculations Scalability of RDBMS and MDDB performance Calculation engine provides full analysis features Source of data transparent to end user

6/19/2012
460
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

460

Architecture Comparison

MOLAP
Definition

ROLAP

HOLAP
Hybrid OLAP = ROLAP + summary in MDDB Sparsity exists only in MDDB part To the necessary extent

MDDB OLAP = Relational OLAP = Transaction level data + Transaction level data + summary in MDDB summary in RDBMS Good Design 3 10 times High (May go beyond control. Estimation is very important) Fast - (Depends upon the size of the MDDB) No Sparsity To the necessary extent

Data explosion due to Sparsity Data explosion due to Summarization Query Execution Speed

Slow

Optimum - If the data is fetched from RDBMS then its like ROLAP otherwise like MOLAP. High: RDBMS + disk space + MDDB Server cost Large transactional data + frequent summary analysis

Cost

Medium: MDDB Server + large disk space cost

Low: Only RDBMS + disk space cost

Where to apply?

Small transactional Very large transactional data + complex model + data & it needs to be frequent summary viewed / sorted analysis

6/19/2012
461
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

461

Representative OLAP Tools:

Oracle Express Products Hyperion Essbase Cognos -PowerPlay Seagate - Holos SAS

Micro Strategy - DSS Agent Informix MetaCube Brio Query Business Objects / Web Intelligence

6/19/2012
462
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

462

Sample OLAP Applications

Sales Analysis Financial Analysis Profitability Analysis Performance Analysis Risk Management Profiling & Segmentation Scorecard Application NPA Management Strategic Planning Customer Relationship Management (CRM)
6/19/2012
463
2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

463

Data Warehouse Testing

464

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Warehouse Testing Overview


There is an exponentially increasing cost associated with finding software defects later in the development lifecycle. In data warehousing, this is compounded because of the additional business costs of using incorrect data to make critical business decisions

The methodology required for testing a Data Warehouse is different from testing a typical transaction system

465

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Difference In Testing Data warehouse and Transaction System


Data warehouse testing is different on the following counts: User-Triggered vs. System triggered Volume of Test Data Possible scenarios/ Test Cases Programming for testing challenge

466

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Difference In Testing Data warehouse and Transaction System.


User-Triggered vs. System triggered

In data Warehouse, most of the testing is system triggered. Most of the production/Source system testing is the processing of individual transactions, which are driven by some input from the users (Application Form, Servicing Request.). There are very few test cycles, which cover the system-triggered scenarios (Like billing, Valuation.)

467

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Difference In Testing Data warehouse and Transaction System


Volume of Test Data The test data in a transaction system is a very small sample of the overall production data. Data Warehouse has typically large test data as one does try to fill-up maximum possible combination of dimensions and facts. Possible scenarios/ Test Cases In case of Data Warehouse, the permutations and combinations one can possibly test is virtually unlimited due to the core objective of Data Warehouse is to allow all possible views of data.

468

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Difference In Testing Data warehouse and Transaction System


Programming for testing challenge In case of transaction systems, users/business analysts typically test the output of the system. In case of data warehouse, most of the 'Data Warehouse data Quality testing' and ETL testing is done at backend by running separate stand-alone scripts. These scripts compare preTransformation to post Transformation of data.

469

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Warehouse Testing Process


Data-Warehouse testing is basically divided into two parts : 'Back-end' testing where the source systems data is compared to the endresult data in Loaded area 'Front-end' testing where the user checks the data by comparing their MIS with the data displayed by the end-user tools like OLAP. Testing phases consists of : Requirements testing Unit testing Integration testing Performance testing Acceptance testing

470

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Requirements testing
The main aim for doing Requirements testing is to check stated requirements for completeness. Requirements can be tested on following factors. Are the requirements Complete? Are the requirements Singular? Are the requirements Ambiguous? Are the requirements Developable? Are the requirements Testable?

471

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Unit Testing
Unit testing for data warehouses is WHITEBOX. It should check the ETL procedures/mappings/jobs and the reports developed. Unit testing the ETL procedures: Whether ETLs are accessing and picking up right data from right source.

All the data transformations are correct according to the business rules and data warehouse is correctly populated with the transformed data.
Testing the rejected records that dont fulfil transformation rules.

472

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Unit Testing
Unit Testing the Report data:

Verify Report data with source: Data present in a data warehouse will be stored at an aggregate level compare to source systems. QA team should verify the granular data stored in data warehouse against the source data available Field level data verification: QA team must understand the linkages for the fields displayed in the report and should trace back and compare that with the source systems Derivation formulae/calculation rules should be verified

473

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Integration Testing
Integration testing will involve following:

Sequence of ETLs jobs in batch. Initial loading of records on data warehouse. Incremental loading of records at a later date to verify the newly inserted or updated data. Testing the rejected records that dont fulfil transformation rules. Error log generation

474

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Performance Testing
Performance Testing should check for : ETL processes completing within time window.

Monitoring and measuring the data quality issues.


Refresh times for standard/complex reports.

475

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Acceptance testing
Here the system is tested with full functionality and is expected to function as in production. At the end of UAT, the system should be acceptable to the client for use in terms of ETL process integrity and business functionality and reporting.

476

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Questions

477

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Thank You

478

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Data Warehouse Concepts

Avinash Kanumuru Diya Jana Debyajit Majumder

2009 Wipro Ltd - Confidential

Content
1 An Overview of Data Warehouse 2 Data Warehouse Architecture 3 Data Modeling for Data Warehouse 4 Overview of Data Cleansing

5 Data Extraction, Transformation, Load

480

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

Content [contd]
6 Metadata Management 7 OLAP 8 Data Warehouse Testing

481

2009 Wipro Ltd - Confidential 2009 Wipro Ltd - Confidential

An Overview
Understanding What is a Data Warehouse

482

2009 Wipro Ltd - Confidential

Вам также может понравиться