Context switching anywhere up to around 20,000 is not really a problem, especially on multi-processor servers, unless you (or your users) actually experience performance degradation in use. If you're seeing numbers that are consistently higher than that over a measured period of time (seconds - ie anything longer than a spike), then you have a problem.
Try increasing the size of the Page File as a preliminary measure - increased Page File activity is often a side-effect of high levels of context-switching, but in extreme cases (set far too low) could actually cause them.
Other things to take into account are;
Hardware - what is the spec of the servers?
Device Drivers - A poorly written driver can cause context switches to go through the roof - check NIC and printer drivers especially.
Users - How many users on each?
Workload - What are typical user loads (what apps do they run, how many, what sort of operations are performed).
Maybe you have a bespoke application which is creating a lot of threads, such as Oracle Financials or similar java-based apps.
The servers are context switching (flipping processor time from one thread to another in swift succession) for a reason. The remedy is not so much to get the number of context switches down by using tweaks, but to discover the root cause and address it. A utility such as CheckIT may help track down hardware issues, and perfmon will allow the monitoring of specific system objects.
RM can be a good friend here - use the Summary Reports to perform a historical analysis - see what has been hitting the server and when. You may need to adjust server load parameters, if you're running MetaFrame XPa or e, or have LB Services installed.
You will find useful tools as ever at
There is also an Advanced Concepts guide available for download from Citrix.com.
I hope this helps CitrixEngineer@yahoo.co.uk
