Performance Counters

[whitneje]

I have an odd behavior with some performance counters I am logging. I have 34 counters on a schedule to log for 2 hours in the morning, every 5 seconds. The odd this is that at the end of the 2 hours I should have 1440 total tests for each counter. ( 12 test/min * 60 min/hr. * 2 hrs. = 1440) The odd thing is, none of the counters logged this many tests. They seem to be grouped into 3 areas. Some logged ~1020, some logged ~660, and some logged ~ 290. There is a correlation to the server performing the test. Why would this be?


Other random question: What is the best way to check CPU usage? CPU usage test, perfcounter, WMI?

[KS-Soft]

The more time server needs to respond the less tests will be performed.
5 sec - interval between end of the check and next check. If average response time 5 sec, HostMonitor will perform about 6 tests per minute (60/(5+5))
BTW: I don't think its a good idea to perform Performance Counter tests every 5 sec, these tests require a lot of resources (compare to TCP or Ping test). Probably you may perform test every 60 sec or once in 3 min?

Other random question: What is the best way to check CPU usage? CPU usage test, perfcounter, WMI?

CPU Usage or WMI. Actually CPU Usage test uses Performance Counters without requesting pdh.dll

Regards
Alex

[thomasschmeidl]

I don't think its a good idea to perform Performance Counter tests every 5 sec, these tests require a lot of resources (compare to TCP or Ping test)

We use a lot of performance counters, that's why I'd like to know whether they require "a lot of resources" on the HM Server or on the target server?

Regards

Thomas

[KS-Soft Europe]

Foregoing Alex's sentence means that it uses a lot of resources when it performs every 5 sec. If you increase time interval, such sentence does not have any sence. Although, as compared with other test methods it really uses a lot of resources. For example, it requires to load additional dlls, like perfproc.dll, perflib.dll, etc. Also we experienced some authentication issues, so we would suggest to use WMI test method, which is much more reliable. Anyway, performance counters test method is very usefull and if you do not have any problems with it, you have not to change it to WMI test method.

Regards,
Max

[KS-Soft]

We use a lot of performance counters, that's why I'd like to know whether they require "a lot of resources" on the HM Server or on the target server?

I think target server uses more resources (CPU and memory) than monitoring system. On the other hand single monitoring system requests many target servers, so total load of monitoring system may be much higher.

Regards
Alex

[thomasschmeidl]

Thanx for the information

Thomas