What I am after is what performance hit is there on the remote machine I am reading a performance counter on? For example, I have a web server that reports some performance counter information on another machine. I want to know what type of performance hit the remote machine takes every time the web server tries to read a performance counter. Does this vary based on which counter you are reading? Is there a standard formula?
The reason I ask is because we have multiple people monitoring the same machine real time using performance counters. I sometimes wonder if we are contributing to the problem and adding extra load on the remote machine we are monitoring.
Thanks.
You'll likely find something in the Sysinternals suite that will allow you to monitor the load caused by counter query routines.
There is no basic formula. Without know what you're querying and where it comes from, I can't begin to guess.
As the previous answer stated, it depends on what you're monitoring and it also depends on how frequently you're monitoring as well. While I can't speak about other tools, the open source tool collectl is something I wrote and which I took great pains to optimize. Even though it is written in perl, it can sample hundreds of counters with < 0.1% overhead at a sampling rate of every 10 seconds. For most users they don't even know it's there and I could easily imaging multiple copies being invisible as well. I'd think other 'native' tools would be even more efficient but might not offer the breadth of functionality collectl does. -mark