Scalability of Office Suites on Terminal Servers

Posted by Benny Tritsch on May 10, 2007

[Introduction] [Methodology] [Performing the Test] [Measurements] [Results] [Conclusions]
[Appendix 1: Configuration Details] [Appendix 2: Step-by-Step Description]

Measurements

Load Generation

A powerful workstation with the vRD tool is used to generate load on each server. The vRD was set up in such a way that a new user can log on to the server via RDP every 30 seconds. It was therefore possible for 120 users to sign on to the server within an hour.

How vRD looked in action can be seen in the following image.

 

 

A configuration file containing the logon information of 120 users was stored to each client and loaded into vRD for the test's execution. A sample configuration file for three users can be seen in the following:

 

<?xml version="1.0" encoding="utf-16"?> <vRDLoadConfigurationFile>   <Testcase ServerName="10.2.2.10">     <Domain>10.2.2.10</Domain>     <User>v1</User>     <Password>OO2007_!</Password>   </Testcase>   <Testcase ServerName="10.2.2.10">     <Domain>10.2.2.10</Domain>     <User>v11</User>     <Password>OO2007_!</Password>   </Testcase>   <Testcase ServerName="10.2.2.10">     <Domain>10.2.2.10</Domain>     <User>v21</User>     <Password>OO2007_!</Password>   </Testcase> </vRDLoadConfigurationFile>

 

The test was be started by pressing the corresponding button. Each new user session could be viewed in its own window in vRD at any time.

Controlling the Servers

Three different documents were opened in the corresponding Office application for each user session during the test. After the users logged on it was necessary to copy the required documents directly to the desktop, if they were not already there. The documents are then opened with the appropriate application. These actions may be executed with a script that is stored in the startup folder.

The logon script on the terminal server with Microsoft Office 2003 can be seen in the following:

 

@echo off if not exist "%userprofile%\Desktop\FinalDocument.doc"   copy c:\Install\Documents\FinalDocument.doc   "%userprofile%\Desktop" if not exist "%userprofile%\Desktop\FinalSpreadsheet.xls"   copy c:\Install\Documents\FinalSpreadsheet.xls   "%userprofile%\Desktop" if not exist "%userprofile%\Desktop\FinalPresentation.ppt"   copy c:\Install\Documents\FinalPresentation.ppt  "%userprofile%\Desktop" echo Document files copied rem Specific extensions here sleep 20 start /B "C:\Program Files\Microsoft Office\Office11\WinWord.exe"   "%userprofile%\Desktop\FinalDocument.doc" echo Launching Word - Documents sleep 20 start /B "C:\Program Files\Microsoft Office\Office11\Excel.exe"   "%userprofile%\Desktop\FinalSpreadsheet.xls" echo Launching Excel - Spreadsheets sleep 20 start /B "C:\Program Files\Microsoft Office\Office11\Powerpnt.exe"   "%userprofile%\Desktop\FinalPresentation.ppt" echo Launching PowerPoint - Presentations

 

Remark: This script was used with only slight modification on the servers with Microsoft Office 2007 or on the 64-bit platforms. Changes generally related to the installation path of the individual office applications. Due to the static nature of the test methodology presented here, a special adaptation was necessary for Office 2007. A registry setting was required to stop Office 2007 applications from blinking on the menu button in the upper left corner after the initial start. Due to the blinking, which also caused unnecessary RDP communication data streams, related system activities were observed, which prevented objective load measurements. Thus, for every terminal server session and before the start of the first accompanying application, Office 2007 had to be brought into a state as if the user had already clicked on a menu item. This was accomplished by setting the registry key OfficeMenuDiscovered under HKCU \SOFTWARE \Microsoft \Office \12.0 \Common \General to the value DWORD:1 via logon script at the position marked “specific extensions here“ (see also appendix).

The documents that were then opened in the log-on script with the related applications had the following characteristics

• FinalDocument.doc opened with Microsoft Word: 3.3 MB
• FinalSpreadsheet.xls opened with Microsoft Excel: 1.2 MB, no embedded macros
• FinalPresentation.ppt opened with Microsoft PowerPoint: 3.4 MB

The following logon script appeared on the terminal server with OpenOffice.org 2.1:

 

@echo off if exist "%HOMEPATH%\Application Data\OpenOffice.org2           \user\config\standard.soc"   goto :JUMP cd "%HOMEPATH%\Application Data" mkdir OpenOffice.org2 xcopy C:\Install\OOO "%HOMEDRIVE%%HOMEPATH%\Application Data" /s /e echo OO Profile created if not exist "%userprofile%\Desktop\FinalSpreadsheet.ods"   copy c:\Install\OOO\Documents\FinalSpreadsheet.ods   "%userprofile%\Desktop" if not exist "%userprofile%\Desktop\FinalDocument.odt"   copy c:\Install\OOO\Documents\FinalDocument.odt   "%userprofile%\Desktop" if not exist "%userprofile%\Desktop\FinalPresentation.odp" copy c:\Install\OOO\Documents\FinalPresentation.odp   "%userprofile%\Desktop" echo Document files copied :JUMP sleep 20 "C:\Program Files\OpenOffice.org 2.1\program\swriter.exe"   "%userprofile%\Desktop\FinalDocument.odt" echo Launching OO Writer - Documents sleep 20 "C:\Program Files\OpenOffice.org 2.1\program\scalc.exe"   "%userprofile%\Desktop\FinalSpreadsheet.ods" echo Launching OO Calc - Spreadsheets sleep 20 "C:\Program Files\OpenOffice.org 2.1\program\simpress.exe"   "%userprofile%\Desktop\FinalPresentation.odp" echo Launching OO Impress - Presentations

 

In order to guarantee the possibility of drawing comparisons between the test results, the exact same documents were used for both Microsoft Office and OpenOffice. The documents used for Microsoft Office were opened with the OpenOffice import filter and saved in the corresponding native OpenOffice formats. These converted documents were then used in the test.

• FinalDocument.odt opened with OpenOffice Writer: 1.4 MB
• FinalSpreadsheet.ods opened with OpenOffice Calc: 290 KB, no embedded macros
• FinalPresentation.odp opened with OpenOffice Impress: 2.5 MB

At this point it was possible to observe that the document formats for OpenOffice generally required less disk space than those for Microsoft Office.

 

Next