December 12, 2013 at 9:52 am
Hi
Can anybody tell me why I'm seeing different values in windows PerfMon Vs SQL Server dmv [dm_os_performance_counters].
For example; when looking specifically at SQLServer:Buffer manager - Free List Stalls/sec, I get a value of 0 in windows PerfMon (great), whereas the dmv gives a constant value of 464 (not great).
Is the dmv accumulating since the last restart of the SQL service? And is the only way to reset these (and others) is by restarting the sql service? Or is there another valid reason why these values are so different?
Thanks
December 12, 2013 at 11:29 am
That counter is an average accumulated over the sample interval (per Books Online and it's counter type). So, the question is, what counts as the interval. So the answer to that is in the other entry in Books Online. /sec counters are cumulative. You need to sample it twice and compare them to see the differences over a sample period.
"The credit belongs to the man who is actually in the arena, whose face is marred by dust and sweat and blood"
- Theodore Roosevelt
Author of:
SQL Server Execution Plans
SQL Server Query Performance Tuning
December 13, 2013 at 3:01 am
Thanks for the reply. Right I get that, so PerfMon is sampling every second and showing the difference (in this case = 0), whereas SQL server DMV is showing the cumulative value (in this case 464) and you need to sample yourself and calculate the difference between last value and current value ( = 0).
Other than a restart of the SQL service, is there anyway to clear / reset the performance counters inside of SQL. I know that dbcc sqlperf can clear latch and wait stats, just wondering about all the others?
Thanks
December 13, 2013 at 3:24 am
I am not aware of a way to reset those counters, no. Sorry.
"The credit belongs to the man who is actually in the arena, whose face is marred by dust and sweat and blood"
- Theodore Roosevelt
Author of:
SQL Server Execution Plans
SQL Server Query Performance Tuning
Viewing 4 posts - 1 through 3 (of 3 total)
You must be logged in to reply to this topic. Login to reply