Tuesday, November 1, 2016

A little bit of play when measuring #SQLServer Resource Governor resource pool reads vs LogicalDisk reads

I use perfmon a lot :-)
So I'm always looking for new patterns.

Today I just happened to graph these two perfmon metrics against each other.  Perfmon was collected with a 30 second interval here.

Its not too surprising that there is a difference.  But how could there be *more* read IOs per second for the default workload group than the total on the server? (See just before 4:00 for a great example.)

Maybe if I sum the two metrics over the 4 hour period they'll even out?

23446639.19   << sum of LogicalDisk reads
22787670        << sum of default pool reads

Huh.  LogicalDisk reads are 2.89% higher than default pool reads.  Not too hard to believe - but for the small stretches where the default pool reads are significantly higher.

Oh well.

Works well for trending.  Not so well if trying to accurately count individual reads.

No comments:

Post a Comment