what is the said to be average that a server should show in the resource manager per processor, looking around through docs i am getting at +- 14000 per processor?
Neil, I took your question to one of our developers and this was his response. FYI - this developer was heavily involved in the coding of the Citrix RMS product (which was originally ours under the SysTrack 1.0 name).
His response:
"That looks mildly high to me. It’s hard to answer without knowing a lot of other things.
The context switch rate will vary more with the number of active threads than with the number of processors in the system. More active threads will necessitate more context switches, but more CPUs to run the threads will result in a lower per-processor rate (since they’ll share the load, and there will be less CPU contention). The context switch rate will also depend heavily on the nature of the application load. It’s actually a lot more complex than this, but perhaps this will help a bit…
If I was forced to guess without any other information being available, I’d say up to 9000 context switches per second on a uniprocessor terminal server is typical. Since the 14K is not radically higher, I’d guess that this is OK. But I would need much more data to be certain."
Was this helpful at all?
Mike
Kapski@yahoo.com
SysTrack 3.1 is available now!
thanks for the input, i am actually getting +- 14k processors on a duel 2.4 Ghz Xeon server (compaq DL) with hyperthreading. Some of the servers are running up to 60,000 when under load. just curious i guess as there is not much documentation on this particular metric, i guess its down to baselining each individual server to see what it can handle
This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
By continuing to use this site, you are consenting to our use of cookies.