Appendix: SQL Server Performance Test Results

This appendix provides test results captured on the performance impact of Microsoft SQL Server 2000 in relation to SQL Server 2005. The following are attributes of this test case:

Test Goals

Test Case Specifics

Test Lab Environment

Test Results

Test Goals

The goal of this test case was to compare the performance impact when using SQL Server 2000 versus SQL Server 2005 as the data store for SharePoint Portal Server 2003 with SP2.

The test bed for the original capacity planning white paper was done a few years ago; therefore it was not possible to retain the same hardware that was used at that time. Consequently, it is not possible to derive direct comparisons between the results of this appendix and the original white paper results.

The new test environment benefited by using products that were not available for the original tests:

  • Each version of SQL Server was installed on a 64-bit version of Windows Server 2003.
  • Visual Studio 2005 was used for stress testing, rather than Application Center Test (ACT).

Top of Page Top of Page

Test Case Specifics

The following section describes the specific tests that were run, in addition to the configurations that were used in the test environment.

Peak Common Action (PCA)

To derive performance results, common portal operations were measured and used as a baseline for the results. The operations used in the previous test case were identical to those tested in this test case. The PCA was weighted as:

  • 50 percent portal home page access
  • 15 percent search operations
  • 15 percent My Site private access
  • 10 percent site directory access
  • 5 percent topic area navigation
  • 5 percent team site access

Farm Configurations

The following section describes the farm configurations and hardware used in the test environment.

Small Farm

  • One Web front-end server computer also running the search and index components
  • One database computer running SQL Server

Medium Farm

  • Two Web front-end servers also running the search component
  • One index server
  • One database computer running SQL Server

Large Farm

  • Eight Web front-end servers
  • Four search servers
  • One index server
  • One database computer running SQL Server

Extra-large Farm

  • Fourteen Web front-end servers
  • Four search servers
  • One index server
  • One database computer running SQL Server

The extra-large farm was configured as described in the original white paper and was tested by using the same operations, but the following tests were added:

  • Topics only
  • Team sites only
  • Home page only

Top of Page Top of Page

Test Lab Environment

The servers used in this test were configured as follows:

  • Web, search, and index servers: dual 2.4 GHz processors with 2 GB RAM.
  • SQL Server: quad 2.2 GHz processor with 16 GB RAM. In addition, this server was running on Windows Server 2003 64-bit Edition.
  • Test Agents: quad 2.2 GHz processors with 16 GB RAM. The test agent computers were running Visual Studio 2005.

All computers were connected via a 1 GB LAN. The traffic on the Web front-end servers was load-balanced by using F5’s Big-IP load balancer. For more information, see F5 Networks - Products - Big-IP.

Top of Page Top of Page

Test Results

The results were captured by measuring the average Requests Per Second (RPS) for each farm size and SQL Server version. The following table presents the performance impact of using SQL Server 2000 versus SQL Server 2005.

Farm size SQL Server 2000
average RPS
SQL Server 2005
average RPS
Change
Small 75 97 +29%
Medium 131 198 +51%
Large 1015 1114 +10%
Extra-large 1638 1932 +18%
Extra-large, topics only 1148 1820 +59%
Extra-large, team sites only 1722 1610 –7%
Extra-large, home page only 1736 1750 +1%

Top of Page Top of Page

 
 
Applies to:
Deployment Center 2003, SPS Admin 2003