VMware Cloud Community
beovax
Enthusiast
Enthusiast

Understanding Disk Latency

This is probably a stupid question but I have been doing some searching and cant really find what I am looking for.

We are doing some fine tuning of our storage array and looking to reduce latency. We havent had any end users complain about perfromance but we are getting the odd alert from veeam monitor and we are not sure if it is a problem or not.

If I take a a typical LUN and look at the perfromance over an hour we would see the following read latency in ms

minium:0

maxium:145

average: 5.4

Looking at the graph we get 3 or 4 spikes every hour above 50ms this is when veeam is flagging a problem.

Does this indicate a problem or do we have the odd workload request which is causing these spikes and nothing to worry about

thanks in advance

0 Kudos
3 Replies
f10
Expert
Expert

I am not sure about the Veem Monitor since I have never used it, but if you are using esxtop from the ESX console and if DAVG/cmd exceeds more than 10ms for a long time then its a latency issue and you need to check the SAN configuration.

For more information on using esxtop to identify storage issues, go through KB the parameters required to check latency are explained very well in the article.

Hope this helps Smiley Happy

If you found this or other information useful, please consider awarding points for "Correct" or "Helpful".

f-10

VCP3, VCP4, HPCP, HP UX CSA

Regards, Arun Pandey VCP 3,4,5 | VCAP-DCA | NCDA | HPUX-CSA | http://highoncloud.blogspot.in/ If you found this or other information useful, please consider awarding points for "Correct" or "Helpful".
beovax
Enthusiast
Enthusiast

thank you for this, we just dug out the "VMware vSphere: Manage for Performance" course material that one of our guys has been on which recommends the same counters.

Looking at the results the average DAVG/KAVG/GAVG/cmd the results are less than 6. Our book says anything below 10 is healthy above 10 needs looking at, above 20 bad news.

But my question is if we record these results to a text file and review them if we see ocassional peaks above 20 maybe say 3-4 per hour is this a problem or just normal activity

0 Kudos
f10
Expert
Expert

If its iSCSI it should be ok, for FC DAVG/cmd should not exceed 15. However since this latency is not for a long time the users are not effected. I would say check everything possible, VM config, application in the VM, FC Switches, Array.

If you found this or other information useful, please consider awarding points for "Correct" or "Helpful".

f-10.

VCP3, VCP4, HPCP, HP UX CSA

Regards, Arun Pandey VCP 3,4,5 | VCAP-DCA | NCDA | HPUX-CSA | http://highoncloud.blogspot.in/ If you found this or other information useful, please consider awarding points for "Correct" or "Helpful".
0 Kudos