Scalability of Office Suites on Terminal Servers

Published in May 2007

The goal of this test was to obtain reproducible and independent information about the scalability of running OpenOffice 2.1 on Terminal Servers compared to running Microsoft Office 2003 and Microsoft Office 2007 on exactly the same platform. The functionality and user-friendliness of any of the Office Suites was regarded as of minor relevance for the findings of this whitepaper. Much more importantly, the results of the test methodology applied was intended to show how many users are able to simultaneously log on to and work with one the respective Office Suites on both 32-bit and 64-bit Terminal Server platforms.

View or download this White Paper from here.

Summary of the Test Results

The results show that about 30 to 40% more users can log on to the platform and launch Microsoft Office 2003 applications as compared to users launching 2.1 applications. Throughout all tests, the hardware was completely identical and the Terminal Services on Windows Server 2003 SP1 were configured the same way. Even for Terminal Servers with Microsoft Office 2007, you can expect that the system resources are used up to 20% more efficiently than it is possible on a system with

For the 32-bit Windows server with 2.1, the hard drive is drawn upon increasingly when there are more than 50 users due to memory used during the testing procedure; this increase in turn leads to longer waiting times for individual users on the system. A similar pattern was observed for the 32-bit server with Microsoft Office 2003, but only with more than 100 users, while the same may be said for Microsoft Office 2007 with 80 users or more.

One of the main reasons behind the differences in performance of the individual Office Suites has to do with the specific memory working set used by the associated applications. The different ways in which page files are used play a significant role for the test results. When the Whindows operating system manages the memory for any of the Microsoft Office product versions in test, data is often paged quite early, whereas this occurs much later with OpenOffice. There are clear differences between the architectures of the Microsoft Office and OpenOffice run-time environments. While Microsoft Office uses standard Windows applications, OpenOffice is based on a container applications serving as control instance and sandbox environment for all OpenOffice applications.

Servers with the 64-bit editions of Windows Server 2003 SP1 were used as a reference for identical tests targeting 2.1, Microsoft Office 2003, and Office 2007. In the 64-bit environment it was possible to confirm the observations of memory and processor resources usage made in the 32-bit environment.