EVA perf - any hints

[Michal Pegierski]

Hello,

I am looking for any metrics hints for EVA perf.
I mean when I can say that I have peformance problems with EVA ?
For example: 20% of CPU utilization is high or not, average queue of 30 is it acceptable or not, etc.

best regards
Michal

[Ruslan]

Hi Michal!

The problems with EVA on host side related to I/O operations. Use perfmon counters such as disk queue length (it must be above 1 in average)

[Michal Pegierski]

Hi Raslan,

I haven't caught your answer. My counters (20%, 30) were examples, not real.
I was expecting general answer.
Is anyone using this tool ?

MichaÅ

[Sheriff Andy]

Michal,

I am not so sure what you need, but we had someone come in and do our performance analysis for us.

Here are some of the metrics that he used;

evaperf hps -cont 300 -csv -dur 604800 > perf_hps.csv (This is for host port statistics)

evaperf cs -cont 300 -csv -dur 604800 > perf_cs.csv (This is for array controller status)

evaperf pdg -cont 300 -csv -dur 604800 > perf_pdg.csv (This is for physical disk groups)

evaperf vd -cont 300 -csv -dur 604800 > perf_vd.csv (This is for virtual disks)

These were all set through scheduled tasks on our SMA. You can get more details from typing in perfview /? at the command line.

Best of luck.

[NathanK]

The only way I would know when something is wrong with our EVA is when the metrics go flat instead peaking and dipping. If your CPU seems to be stuck at XX% for a period of time or your port requests won't go higher than XXK then you know that you've reached a bottleneck somewhere.