PA and protocol latency reporting

[jkennedy_4]

What does the "Average Latency per Protocol" graph/stat truly indicate in PA for NFS?

Is it latency for every NFS op that it's satisfying?  Some of them?  The highest latent operation?

I don't fully understand it because in PA, and running sysstat, there can be things going on that all seem to be happening very quickly, yet latency is huge.  I mean in the tens of thousands of microseconds.

The particular case in question is attached as png (hopefully).

Thanks.

[aborzenkov]

10 thousands of microseconds is not huge. It is just 10ms, which is not spectacular, but far below "huge".

[jkennedy_4]

If it were only 10k microseconds, maybe. But note the png I attached to the post, it shows 25k-35k microseconds. 25-35 milliseconds is pretty darned huge.

Jeff Kennedy

Qualcomm, Incorporated

QCT Engineering Compute

858-651-6592