sql server 2012 analysis services partitioning performance demonstration - print version

6
05/12/2013 SQL Server 2012 Analysis Services Partitioning Performance Demonstration - Print Version www.mssqltips.com/tipprint.asp?tip=3012 1/6 source: http://www.MSSQLTips.com/tip.asp?id=3012 -- printed: 12/5/2013 1:37:27 PM SQL Server 2012 Analysis Services Partitioning Performance Demonstration Written By: Dallas Snider -- 8/1/2013 Problem How much of an improvement in processing time can a simple SQL Server Analysis Services (SSAS) partitioning structure provide? Can you provide any examples? Does this processing time improvement only apply to the fact tables? Check out this tip to learn more. Solution In this tip, we will show the differences in processing times between a cube with the default single partition created by Visual Studio and a cube utilizing partitions based on the year of an event in a fact table. Also, we will see how the processing times change with increasing numbers of records in the fact table. The cube structure for both of the cubes in this tip is shown below. There was one measure group containing three measures with each measure having four dimensions. The only difference between the two cubes was the partitioning structure.

Upload: abacusdotcom2964

Post on 22-Oct-2015

24 views

Category:

Documents


0 download

DESCRIPTION

SQL Server 2012 Analysis Services Partitioning Performance Demonstration

TRANSCRIPT

Page 1: SQL Server 2012 Analysis Services Partitioning Performance Demonstration - Print Version

05/12/2013 SQL Server 2012 Analysis Services Partitioning Performance Demonstration - Print Version

www.mssqltips.com/tipprint.asp?tip=3012 1/6

source: http://www.MSSQLTips.com/tip.asp?id=3012 -- printed: 12/5/2013 1:37:27 PM

SQL Server 2012 Analysis Services PartitioningPerformance Demonstration

Written By: Dallas Snider -- 8/1/2013

Problem

How much of an improvement in processing time can a simple SQL Server Analysis Services (SSAS) partitioningstructure provide? Can you provide any examples? Does this processing time improvement only apply to the facttables? Check out this tip to learn more.

Solution

In this tip, we will show the differences in processing times between a cube with the default single partition createdby Visual Studio and a cube utilizing partitions based on the year of an event in a fact table. Also, we will see howthe processing times change with increasing numbers of records in the fact table.

The cube structure for both of the cubes in this tip is shown below. There was one measure group containing threemeasures with each measure having four dimensions. The only difference between the two cubes was thepartitioning structure.

Page 2: SQL Server 2012 Analysis Services Partitioning Performance Demonstration - Print Version

05/12/2013 SQL Server 2012 Analysis Services Partitioning Performance Demonstration - Print Version

www.mssqltips.com/tipprint.asp?tip=3012 2/6

The Partitions tab of the cube with the default single partition is shown here. Note that by default, the EstimatedRows is set to zero.

The Partition Wizard was used to create the six partitions for this tip for each calendar year from 2005 through 2010inclusive. For each partition, the WHERE clause was modified to restrict the rows returned within a one calendar yearrange.

Page 3: SQL Server 2012 Analysis Services Partitioning Performance Demonstration - Print Version

05/12/2013 SQL Server 2012 Analysis Services Partitioning Performance Demonstration - Print Version

www.mssqltips.com/tipprint.asp?tip=3012 3/6

After the Partition Wizard completed, the Slice property for each partition was set to its corresponding tuple for thecalendar year.

Page 4: SQL Server 2012 Analysis Services Partitioning Performance Demonstration - Print Version

05/12/2013 SQL Server 2012 Analysis Services Partitioning Performance Demonstration - Print Version

www.mssqltips.com/tipprint.asp?tip=3012 4/6

On the Aggregations tab, one aggregation was created using the Aggregation Design Wizard. The wizard was set todesign aggregations until the performance gain reached 30 percent. Because the fact table data was distributedequally among the year values, the one aggregation was assigned for all partitions.

After creating one partition for each year and the aggregation scheme, the Partitions tab appears as shown below.

The populations of the dimensions remained consistent throughout the performance tests and are shown here.

Customer dimension: 1,000,000 recordsDate dimension: 2,556 records representing January 1, 2005 through December 31, 2010Facility dimension: 1,000 recordsProduct dimension: 606 records

Twenty million records were inserted into the fact table. The cube with 6 partitions was processed 3 times with the"Process Full" option and its processing time was averaged. Next, the cube with the default single partition wasprocessed 3 times with the "Process Full" option and its processing time was averaged. This processing was repeatedwith fact tables of 40, 60 and 80 million records.

The results are shown below in tabular and graphical form. As the number of records in the fact table increases, wecan see where the processing time for the single partition cube increases at a greater rate than the six partitioncube. Thus the importance of partitioning becomes more critical.

Page 5: SQL Server 2012 Analysis Services Partitioning Performance Demonstration - Print Version

05/12/2013 SQL Server 2012 Analysis Services Partitioning Performance Demonstration - Print Version

www.mssqltips.com/tipprint.asp?tip=3012 5/6

Next Steps

Try different partitioning and aggregation strategies to see how much your processing time is reduced.Please note that partitioning can reduce the processing time, but not necessarily improve the query responsetime.Please refer to the following tips for further assistance with SQL Server Analysis Services:

SQL Server Analysis Services Partition Maintenance and ProcessingSQL Server Analysis Services (SSAS) TutorialAll SQL Server Business Intelligence Tips

Copyright (c) 2006-2013 Edgewood Solutions, LLC All rights reservedprivacy | disclaimer | copyright | advertise | about

authors | contribute | feedback | giveaways | user groups | whitepapers Some names and products listed are the registered trademarks of their respective owners.

Page 6: SQL Server 2012 Analysis Services Partitioning Performance Demonstration - Print Version

05/12/2013 SQL Server 2012 Analysis Services Partitioning Performance Demonstration - Print Version

www.mssqltips.com/tipprint.asp?tip=3012 6/6