Server Fault is a question and answer site for system and network administrators. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Assume two servers, A and B. From server A I open perfmon and add the counter \ServerB\Processor\% Processor Time and start monitoring it every second.

  1. Is Server A going to poll server B every second for CPU data (I guess this is the case)
  2. Is Server A going to suffer some % of performance degradation due to this constant polling process (it may not be an issue if you monitor 1 counter, but you can imagine monitoring 100's of them in a performance test)
  3. Is Server B going to suffer some % of performance degradation, due to multiple consumers (like Server A) constantly polling for data.
share|improve this question

migrated from Aug 12 '11 at 13:30

This question came from our site for professional and enthusiast programmers.

1-Yes:I am not aware of any eventing interface behind P[erformance]C[ounter], this is just RPC.

2-This really depends on several different things:

  • distance [network latency]
  • nr. of computers
  • nr. of counter for each computer

    For example:I was not able to pull 1000PC/s with the Net API, even using Net4 [TPL] and a producer/consumer model. I am able to do this via WMI event queries.

3-Don't know, habe not tracked this intensively, but does not found a strong indicator for it. Using WMI - my perf app above - it causes something between 1-2% of CPU on each monitored server [using 15PC/5s].

Hope, this helps.


share|improve this answer

...but you can imagine monitoring 100's of them in a performance test...

In a performance test our goal is to find representative values and not sample so high as to actually cause a distortion in performance. Depending upon the length of the test the sample window my be once every 5 seconds, 10 seconds, on up to 30 seconds, but only at the every second mark if using a very small number of top level metrics (SYSTEM|APP of CPU, Disk Bytes, Network Bytes, RAM Committed pool) and then only when doing some engineering discovery in follow up to another issue.

There is definitely a quantum effect on monitors! The more more aggressive the number and shorter the frequency between samples then the act of pulling the values becomes a distortion on what you are trying to sample.

share|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.