views:

354

answers:

2

What limits exist on the amount of data one can publish to a custom Windows performance counter category?

I understand there is no hard limit on the number of counters or the number of instances, but rather there is a memory limit for the entire category. What is that limit?

Is there a limit on the total number or size of all performance counter categories? What else should be taken into account when dealing with a relatively large amount of data that needs to be published?

To put this into perspective, I need to publish around 50,000 32bit counter-instance-values. I could split these up into categories in various ways, depending on what limits exist.

I appreciate that performance counters may not be the best solution, but there are reasons for this madness.

A: 

Is there actually a limit though? I thought you basically just published a block of shared memory - why not just increase the size of the block? What makes you think there is a limit?

1800 INFORMATION
+2  A: 

Under what circumstances would you need to publish tens of thousands of counters.

Remember that the tools that read those perf counters typically aren't designed for such massive data sets (althought they might be). As a result, it is possible that while you'll be able to author such a data set, the tools that read your data will fail in "interesting" ways.

You might want to reconsider your need to collect so much data. Do you really need 50,000 perf counters? What will you do with the information once you collect it? Will you really be able to gather meaningful information from 50,000 counters?

Larry Osterman
Yes, I really need to publish all those counters, and the tool that will be reading them can handle it. Thanks for your response but you have not made any attenpt to answer the question. -1
Matt Howells
I am not the original poster, I have an app that can publish an essentially indefinite list of counter instances - the exact number is more or less data driven. The counter instances aggregate the same data across different categories and criteria - this is quite useful for monitoring
1800 INFORMATION