Difference in Volume Latency for Cifs Latency by Optype vs Avg Latency Per Protocol at System Level

[JGPSHNTAP]

    Ok fellow DFM PA users, I need some clarification b/c i'm coming up empty a little..

We are seeing two different latency numbers when I'm viewing a particular issue

Under the system storage summary view in PA, if we look at Avg Latency by per protocol, we see an issue with cifs starting lets say 1.5 hours earlier, then if we were looking at the volume level latency for cifs by optype.

So, what is the difference between those two views and which is more reliable for troubleshooting cifs latency.

When I look at it on the filer level i usually do something like

stats show -n 50 -i 1 volume:*:cifs_read_latency volume:*:cifs_write_latency

I think the one in system summary view is -  stats show -n 50 -i 1 cifs:*:cifs_latency

Any explaination is greatly appreciated it.

Thanks    

[rmatsumoto]

Avg latency by per protocol is coming from the system counter group and the latter is coming from the volume counter group.  It's been more than a year since we had the issues, but for us, the case where we rely on the volume:cifs_* counters is when there's an AV performance issue.  I'm actually not sure if that counter gave me a different number than system:cifs_latency, but we could very clearly tell that there was a correlation between our AV server performance and volume :cifs_*latency values.  The volume:cifs_latency(or cifs_read_latency... one of these) got high while the volume:read_latency and volume:avg_latency stayed low.