Thursday, February 6, 2014

Build, Deploy and Process of SSAS cube

This blog is all about Build, Deploy and Process of SSAS cube.

Firstly, when we build the cube, the build process will create four XML files inside the bin
folder. As the names hint at, the asdatabase file is the main object definition file and the
three other files contain configuration and deployment metadata.

Secondly, the deployment process will attach the above files to running Server Analysis
Service. This deployment can be done by many ways, but more appropriate option for
production deployment is deploy by Deployment Wizard Tool, because this tool will allow
doing many configuration settings.

The following screenshots show the steps of deployment via Deployment Wizard Tool.
If necessary we can create script as well. This will create XMLA file, which contains all the
configurations and related data. The XMLA is stand for “XML for Analysis” open standard supports
data access to data source that reside on the World Wide Web. Microsoft SQL Server Analysis Services
implements XMLA per the XMLA 1.1 specification. XMLA is a Simple Object Access protocol (SOAP) –
based XML protocol, designed for universal data access to any standard multidimensional data source
residing on the web.
Finally, the Processing is the step, or series of steps, in which Analysis Service loads data
from a relational data source into a multidimensional model. For objects that use MOLAP
storage, data is saved on disk in the database file folder. For ROLAP storage, processing
occur on demand, in response to MDX query on an object. For object that uses ROLAP
storage, processing refers to updating the cache before returning query results.

Processing Objects: Processing affects the following Analysis Service objects: measure
groups, partitions, dimensions, cubes, mining models, mining structure, and databases.
When an object contains one or more objects, processing the highest-level object causes a
cascade of processing all the lower-level objects. For example, a cube typically contains
measure groups (each of which contains one or more partitions) and
dimensions. Processing a cube causes processing of all the measure groups within the
cube and the constituent dimensions that are currently in an unprocessed state.

While the processing job is working, the affected Analysis Service Objects can be accessed
for querying. The processing job works inside a transaction and the transaction can be
committed, or rolled back. If the processing job fails, the transaction is rolled back. If the
processing job succeeds, an exclusive lock is put on the object when changes are being
committed, which means the object is temporarily unavailable for querying or processing.
During the committed phase of the transaction, queries can still be sent to the object, but
they will be queued until the commit is completed.
Processing stage can be check using the object properties, the below image show the Cube
processing stage.

Likewise we can check object wise processing stage. For some object if properties option is
not available, then we can use the report for this.
When you process objects in Analysis Services, you can select a processing option, or you
can enable Analysis Services to determine the appropriate type of processing. The
processing methods made available differ from one object to another, and are based on the
type of object. Additionally, the methods are based on what changes have occurred to the
object since it was last processed. If you enable Analysis Service to automatically
select a process method (Default), it will use the method that returns the object to a
fully processed stage in the least time.

Processing Mode: The following table describes the processing methods that are available
in Analysis Services, and identifies the objects for which each method is supported.
Mode Applies to Description

Process Cubes, databases, Detects the process state of database objects, and
Default dimensions, measure performs processing necessary to deliver
groups, mining unprocessed or partially processed objects to a
models, mining fully processed state. If you change a data binding,
structures, and Process Default will do a Process Full on the
partitions. affected object.

Process Cubes, databases, Processes an Analysis Services object and all the
Full dimensions, measure objects that it contains. When Process Full is
groups, mining executed against an object that has already been
models, mining processed, Analysis Services drops all data in the
structures, and object, and then processes the object. This kind of
partitions. processing is required when a structural change
has been made to an object, for example, when an
attribute hierarchy is added, deleted, or renamed.
 Process Cubes, databases, Drops the data in the object specified and any
Clear dimensions, measure lower-level constituent objects. After the data is
groups, mining dropped, it is not reloaded.
models, mining
structures, and
partitions.

Process Dimensions, cubes, Processes data only without building aggregations

Data measure groups, and or indexes. If there is data is in the partitions, it will
partitions. be dropped before re-populating the partition with
source data.

Process Dimensions, measure For dimensions, adds new members and updates
Add groups, and partitions dimension attribute captions and descriptions.
For measure groups and partitions, adds newly
available fact data and process only to the relevant
partitions.

Process Dimensions Forces a re-read of data and an update of

Update dimension attributes. Flexible aggregations and
indexes on related partitions will be dropped.

Process Cubes, dimensions, Creates or rebuilds indexes and aggregations for

Index measure groups, and all processed partitions. For unprocessed objects,
partitions this option generates an error.
Processing with this option is needed if you turn off
Lazy Processing.

Process Cubes and mining If the cube is unprocessed, Analysis Services will
Structure structures process, if it is necessary, all the cube's
dimensions. After that, Analysis Services will
create only cube definitions. If this option is
applied to a mining structure, it populates the
mining structure with source data. The difference
between this option and the Process Full option is
that this option does not iterate the processing
down to the mining models themselves.

Process Mining structures Removes all training data from a mining structure.
Clear
Structure
Source: http://msdn.microsoft.com/
Processing Settings : The following table describes the processing settings that are
available for use when you create a process operation.
Processing Description
Option

Parallel Used for batch processing. This setting causes Analysis Services to fork off
processing tasks to run in parallel inside a single transaction. If there is a failure,
the result is a roll-back of all changes. You can set the maximum number of
parallel tasks explicitly, or let the server decide the optimal distribution. The
Parallel option is useful for speeding up processing.

Sequential Controls the execution behavior of the processing job. Two options are available:
(Transaction · One Transaction. The processing job runs as a transaction. If all processes
Mode) inside the processing job succeed, all changes by the processing job are
committed. If one process fails, all changes by the processing job are rolled back.
One Transaction is the default value.
· Separate Transactions. Each process in the processing job runs as a stand-
alone job. If one process fails, only that process is rolled back and the processing
job continues. Each job commits all process changes at the end of the job.
When you process using One Transaction, all changes are committed after the
processing job succeeds. This means that all Analysis Services objects affected by
a particular processing job remain available for queries until the commit process.
This makes the objects temporarily unavailable. Using Separate
Transactionscauses all objects that are affected by a process in processing job to
be taken unavailable for queries as soon as that process succeeds.

Writeback Controls how writeback tables are handled during processing. This option applies
Table Option to writeback partitions in a cube, and uses the following options:
· Use Existing. Uses the existing writeback table. This is default value.
· Create. Creates a new writeback table and causes the process to fail if one
already exists.
· Create Always. Creates a new writeback table even if one already exists. An
existing table is deleted and replaced.

Process Controls the object scope of the processing job. An affected object is defined by
Affected object dependency. For example, partitions are dependent on the dimensions that
Objects determine aggregation, but dimensions are not dependent on partitions. You can
use the following options:
· False. The job processes the objects explicitly named in the job and all
dependent objects. For example, if the processing job contains only dimensions,
Analysis Services processes just those objects explicitly identified in the job. If the
processing job contains partitions, partition processing automatically invokes
processing of affected dimensions. False is the default setting.
· True. The job processes the objects explicitly named in the job, all dependent
objects, and all objects affected by the objects being processed without changing
the state of the affected objects. For example, if the processing job contains only
dimensions, Analysis Services also processes all partitions affected by the
dimension processing for partitions that are currently in a processed state. Affected
partitions that are currently in an unprocessed state are not processed. However,
because partitions are dependent on dimensions, if the processing job contains
only partitions, partition processing automatically invokes processing of affected
dimensions, even when the dimension is currently in an unprocessed state.
 Dimension Key Determines the action taken by Analysis Services when errors occur during
Errors processing. When you select Use custom error configuration, you can select
values for the following actions to control error-handling behavior:
· Key error action. If a key value does not yet exist in a record, one of these
actions is selected to occur:
o Convert to unknown. The key is interpreted as an unknown member. This is
the default setting.
o Discard record. The record is discarded.
· Processing error limit. Controls the number of errors processed by selecting
one of these options:
o Ignore errors count. This will enable processing to continue regardless of the
number of errors.
o Stop on error. With this option, you control two additional settings. Number of
errors lets you limit processing to the occurrence of a specific number of errors.
On error action lets you determine the action when Number of errors is reached.
You can select Stop processing, which causes the processing job to fail and roll
back any changes, or Stop logging, which enables processing to continue without
logging errors. Stop on erroris the default setting with Number of errors set to 0
and On error action set to Stop processing.
· Specific error conditions. You can set the following options to control specific
error-handling behavior:
o Key not found. Occurs when a key value exists in a partition but does not exist
in the corresponding dimension. The default setting is Report and continue.
Other settings are Ignore error and Report and stop.
o Duplicate key. Occurs when more than one key value exists in a dimension.
The default setting is Ignore error. Other settings are Report and continue and
Report and stop.
o Null key converted to unknown. Occurs when a key value is null and the Key
error action is set to Convert to unknown. The default setting isIgnore error.
Other settings are Report and continue and Report and stop.
o Null key not allowed. Occurs when Key error action is set to Discard record.
The default setting is Report and continue. Other settings areIgnore error and
Report and stop.
When you select Use default error configuration, Analysis Services uses the error
configuration that is set for each object being processed. If an object is set to use
default configuration settings, Analysis Services uses the default settings that are
listed for each option.
Source: http://msdn.microsoft.com/
Cheers!