Installation guide

24 Implementing Windows Terminal Server and Citrix MetaFrame on IBM ^ xSeries Servers
You can find IBM benchmarks for Windows Server 2003 and Citrix on the Web at:
http://www.pc.ibm.com/ww/eserver/xseries/benchmarks/series.html
3.2.5 Using script results to assess real-world requirements
An underestimation can occur due to the differences between the benchmark environment in
which WTS scripts are run and the real world environment. Scripts generate a synthetic
workload that provides repeatability but that is not particularly representative of any specific
real-world situation. Repeatability allows comparison of the effects of changes to the server
and software configuration. This means that the server may be finely tuned for this specific
and stable workload. Production environments are more subject to variation than this.
Another important factor is the rate at which the workload is placed on the server. In
benchmarks, some effort is made at randomizing the work presented to the server, but real
users’ work patterns can create significant peaks and valleys on a daily, weekly, or even
longer cyclical basis. For example, daily peaks can be expected when everyone logs on to the
system when they arrive in the morning, while workload dips usually occur at lunch time and
overnight.
In practice, test results provide appropriate “rules of thumb” for most users who run
applications such as Microsoft Office, IBM Lotus Notes, or Microsoft Outlook. For more
complicated or custom applications, you should be more conservative in your server sizing
and use the results from the structured task workers, power users, or heavy user scripts.
For this situation, IBM and other vendors have developed guidelines to map Terminal Server
script results to recommended maximum values that customers can use for production
servers for particular client workloads.
The following rule of thumb determines the upper bound on the number of production WTS
clients actively connected at any one time to your xSeries server:
80% of script result = maximum number of production WTS users
For example, a particular script may be running on a particular server configuration and
indicate that a maximum of 100 terminal server users are supported. In this case, you can
expect to support up to about 80 users of the type represented by the test script in your
production environment.
This rule of thumb provides some spare capacity for your production environment. These
guidelines vary from vendor to vendor. However, they all aim to provide a reasonable safety
margin in their calculations. With this, you can feel confident that your server can support the
load that is quoted and cope with reasonable peaks in demand.
Use the following rule of thumb as a useful guide to provide a minimum estimate for the
amount of RAM required for your terminal server if you have not conducted a pilot. The
minimum amount of RAM you need in your terminal server is:
128 + 15 x (Concurrent number of terminal server users) MB
Tip: Up to 90 concurrent knowledge workers or medium users can be supported on a dual
Intel Xeon processor terminal server and receive acceptable response times. In practice,
this relates to general application use of such products as Microsoft Office and Lotus
Notes. Using our rule of thumb, this equates to 90 x 0.8 = 72 concurrent production users
or approximately 70 users.
Rule of thumb
for WTS users
Rule of thumb
for terminal
server RAM