Вы находитесь на странице: 1из 25

Database Platform Discussion

Option 1

IBM Solution

Main Data Centers

DR Center

DB replications

Site A

A
B1(idle)

site B

A1(idle)

A2 (idle)

B2 (idle)

Pros
Built-in DR engine without additional components
Simple architecture, no data consistency handling across sites

Cons
Required double storage spaces
Standby supports only read only or idle configuration
Data need to be partitioned for site A and site B

IBM Solution

Option 2
Geographically Dispersed Cluster (GDPC)

IBM Solution

Geographically Dispersed Cluster + database replication


Site 1

Site 2

DR

(Active)

(Active)

(Passive)

M1, M2

M3, M4

M1, M2

M3, M4

Database
Replication
HADR

Pros
Database handles all data synchronization across site1, site2 and DR
Single database

Cons
Require dark fibre or 10GB Ethernet between site 1 and site 2
Require additional site ( site3 ) as judge to split brain

Reference of using GDPC

IBM Solution

BCOM
- Internet Payment System, Inter-banks payment systems
- Cross site cluster

Dark Fiber or 10 Gb WAN

Infiniband

Mbr
1

Mbr
3

N kms

CF
prim

site
tie-breaker

Ethernet

Log

CF
sec

3rd

Log

GPFS Shared Storage (logical view)

Mbr
2

Mbr
4
Fiber Channel

Log

Log

Database

Physical Storage site 'A'

GPFS replication between sites

Physical Storage site 'B'

IBM Solution

Reference of using virtualization

BTMU
- Retail Banking Application
- 1TB compressed database size
- 3 years production

2core
84GB Memory

2core
84GB Memory

Reference of using DB2 with high transaction requirement


IBM Solution

HKEx server environment


HP Server and Storage in customer site is used
HP DL360 G5 Server

SAN Storage server

Two Intel Quad-Core E5450 Processors

2 HP VRAIDs from 8 physical disk drives

12GB RAM

VRAID 0 for Transaction log file system

VRAID0

locally-attached 145GB (SAS + RAID1)

VRAID 1 for database file system

VRAID1

Red Hat Enterprise Linux 5.3


*

RSS1

RSS2

DB2 achieved
30% more transaction
than Oracle per min
On same hardware
7

LUN1-A

LUN1-A

LUN1-A

LUN1-A

LUN0

LUN0

LUN0

LUN0

LUN1-B

LUN1-B

LUN1-B

LUN1-B

LUN0

LUN0

LUN0

LUN0

* This disk layout is an assumption based on the information


provided by the cust :
2 HP VRAIDs( VRAID0 and VRAID1) from 8 physical disk
drives.

IBM Solution

Longevity in Transaction Processing Performance


Benchmarks are often a game of leapfrog
DB2 has demonstrated sustained superiority over competitors

Results as of Nov 20, 2014

IBM Solution

Simple Package with all Enterprise level functions

Features
BLU
pureScale
Data Partitioning Feature
Optim Performance Manager Extended Edition
Optim Query Workload Tuner
Optim Configuration Manager
InfoSphere Optim pureQuery Runtime for LUW
Warehouse Model Packs
WHS SQW
WHS Mining & text Analytics
Cognos 10.2 5 Users
WHS Cubing Services
InfoSphere Data Architect Capabilities
solidDB & solidDB UC
IBM Mobile Database Sync
pureXML Storage
Time Travel Query
Compression
Row & Column Access Control
Label Based Access Control
SQL Replication
Federation Capabilities
Websphere MQ
High Availability Disaster Recovery
Tivoli System Automation
Multi-temperature Storage
Materialized Query Tables
Multi-Dimensional Clustering
Table Partitioning
Scan Sharing
DB2 Workload Management (WLM)
MQ Replication / CDC

Desription
In-memory columnar capabilities
Active-Active
Database partitioning
Performance monitoring
Workload monitoring
Configuration Admin
Query optimization
Data Warehouse Model
ETL tool
Data Minining
BI front end
BI Cube
Data Modeler
In memory database cache
Mobile database
XML database
Database auditing
Compression
Security
Security
Replication
DB Feberation cross database
Message Queue
Log shipping
Clustering software
Hot/Warm/Cold multi-tierdata storage
MQT
Storage clustering
Table Partition
Scan result sharing
Query workload priorization
CDC

Similar to Oracle Feature


Exadata - storage IO
RAC
No offer
Tuning Pack
Workload manager
Configuration Pack
No offer
No offer
Data Integrator
Data Minining
OBIEE
OBIEE Cube
Designer
Times Ten
Mobile Server
XML type
Auditing expert
Adv. Compression
Adv. Security
Label Security
DB Link
Transparent Application Gateway
MessageQ
Data Guard
Veritas
No offer
MQV
No offer
Partioning Feature
Exadata - scan sharing
Workload manager
Golden Gate

IBM Solution

Virtualization Licensing
With SQL Server you have two options
1. License all virtual cores for a VM (min 4 core licenses) can end up licensing more
cores than exist on a machine
2. License all physical cores on a machine (Enterprise Edition only)
With DB2
Sub-Capacity Licensing allows for licensing of only the physical cores used by DB2
With Workgroup can run unlimited VMs of DB2 on up to 16 sockets

10

IBM Solution

Server options in DB2

HA pair

Small applications

Share disk + cluster software

Database replication pairs


Active

Idle
HA pair

DR or remote site
requirement

IBM Solution

Server options in DB2

Concurrent Access Cluster


Data node

Data node

Data node

PPRC/Q-rep

Concurrent Access Cluster


Data node

Data node

Continous availbility
SAN

SAN

Support rolling fixpak upgrade and non-interrupt member addition in v10


Support PPRC ( active-passive ) and Q-rep ( active-active )
Extended Purescale provides cross site clustering within 30Km

Data node

IBM Solution

Roadmap for RDBMS

IBM Solution

DB2 with BLU Acceleration


Rich capability integrated with IBM DB2 10.5
Fast Answers. Simply Delivered.

What is DB2 with BLU Acceleration?


In-memory analytic database
Multiple IBM innovations
In-memory processing
Columnar storage
Multiplatform: Supports AIX, Linux, zLinux, Windows
Ready for Analytics with no tuning effort
Analyze more data faster and more
efficiently

IBM Solution

Server Options (New Cloud option)

Guesswork, Overhead, Risk

Data

Admin

Software

Roll-Your-Own
Hardware

15

Data

Admin

No Guesswork
Smallest Risk
Guaranteed Uptime

No Tuning
Truly Agile
Appliance Simplicity

Data
Layer

Data
Warehouse

Database

Data Warehouse

as a Service

as a Service

Software

Hosted

2014 IBM Corporation

IBM Solution

Performance

IBM Solution

10 times
performance
increase to 30,000
orders per second

IBM Solution

Benchmark Result
Goal : Achieve not less than 30k/sec with 100000000 records
Result from IBM : 97k/sec

Thread 8 completed, wrote 12501000 recs in 1024354ms


Thread 3 completed, wrote 12501000 recs in 1025968ms
Thread 2 completed, wrote 12501000 recs in 1026454ms
Thread 1 completed, wrote 12501000 recs in 1027078ms
Thread 7 completed, wrote 12501000 recs in 1026548ms
Thread 4 completed, wrote 12501000 recs in 1030068ms
Thread 6 completed, wrote 12501000 recs in 1030140ms
Summary: W 8 3000 (O=22224000 U=22224000 Q=22224000 T=11112000
Z=44448000) 100008000 1030140 97000

IBM Solution

Row compression Performance


With the following compression ratio, we find overall performance from
65k/sec to 97k/sec

Table
Compression

Ratio for HKEx testing tables

No Comp
(MB)

Rows

Comp
(MB)

Ratio
(%)

tblOrder_cm

22,224,000

5,428

880

84

tblTlog1_cm

5,556,000

1,551

225

85

tblTlog2_cm

5,556,000

1,551

225

85

tblTlog3_cm

5,556,000

1,551

225

85

tblTlog4_cm

5,556,000

1,551

225

85

tblTlog5_cm

5,556,000

1,551

225

85

tblTlog6_cm

5,556,000

1,551

225

85

tblTlog7_cm

5,556,000

1,551

225

85

tblTlog8_cm

5,556,000

1,551

225

85

tblTrade_cm

11,112,000

2,413

510

79

Total

77,784,000

20,249

3,190

84

IBM Solution

Database Management

IBM Solution

Deep Compression
Adaptive Compression is an enhancement to the Classic Row Compression
Benefits
*60% storage space savings over Classic
Row Compression
TABLE
Reduce I/O
Static table
DB2 page
Faster queries
level
Dynamic page
level dictionary

DB2 page
Dynamic page
level dictionary

Better compression ratios than Classic Row Compression


Over time reduces need of a REORG as page-level dictionaries
adapt to data skew over time

dictionary

DB2 page
Dynamic page
level dictionary

IBM Solution

Hybrid Engine for native XML data support


DB SERVER
CLIENT

SQL/XML

Relational
Interface

DB2 Client /
Client
Application

DB2 Storage:
Hybrid

Relational

DB2
Engine

XQuery

XML
Interface

XML

IBM Solution

Hybrid Engine for native multiple language

DB2 Server

SQL PL

DB2
Compiler

PL/SQL

PL/SQL
Compiler

DB2 Engine

IBM Solution

Workload Management
Map statement to service class (Predictive)
SQL compiler predicts what data (and table spaces) will be touched by the SQL
statement and builds the list of data tags
List of data tag will define the initial service class placement
Tablespace
Hot

Superclass MAINSC
1,2,3
Default
Workload

Work
action set

4,5,6
7,8,
9

Subclass
SCHIGH
Subclass
SCMED

Tablespace

Warm

Subclass
SCLOW

Tablespace
Cold

IBM Solution

Manageability of Aging Data


Partitioned Table Sales

Partition
Automatic
Storage
Table
space

2012Q1

Table Space 14

Storage
Group

2011Q3

2011Q2

Table Space 13
Table Space 12Table Space 11

2011Q1 2010Q4 2006Q3

Table Space 10 Table Space Table Space 1


9

spath: /hot/fs1

New

2011Q4

SG_HOT

spath: /warm/fs1
spath: /warm/fs2

SG_WARM

spath: /cold/fs1
spath: /cold/fs2
spath: /cold/fs3

SG_COLD

Physical Disk
SSD RAID Array

FC/SAS RAID Array

SATA RAID Array

Вам также может понравиться