Does anyone have a document showing acceptable ranges for storage statistics for esxtop? In particular, i'm trying to pin down disk latency in Exchange 2003 luns. I show CMDS/s in the hundreds, and DAVG/ cmd averaging around 15 or so. I know this is high, but can't find any documented acceptable ranges. Most other vm's are always at zero. Thanks.
Have you expanded out the LUN statistics for each adapter? That will give you a view of each LUN vs the vmhbaX
e Expand/Rollup Disk Statistics
a Expand/Rollup Disk Channel Statistics
t Expand/Rollup Disk Target Statistics
l Expand/Rollup Disk Lun Statistics
QUED and ACTV is interesting to look at too.
Anyhow, 15ms is .... OK. It's not terrible and not terrific. You probably also need to take a look at global SAN latency to see if the SAN disks are the bottleneck. If so, you need faster disks or move your Exchange data to it's own disk spindles.
Ben
To start I always look at the QUED field for disk bottlenecks. 10 is slow anything higher and it's bad. If everything looks ok on the host, check the counters on the SAN. That's where you will get your real information.
What SAN do you have?
Thanks, I've expanded it all out various ways, found originally every datastore/lun was going out the same default path (vmhba1). Changing the Exchange luns to path 2 helped decrease the latency greatly, but I still want to verify the numbers I am getting.
Thanks, my QUED is consistently zero, so at least that's going my way. We use an HP XP10000 san. I'm basically concerned bout the ridiculous number of commands per second on Exchange luns, and then latency per command. If you do the math, that adds up.
At this point you should use your SAN monitoring software to view the performance of the lun. I'm not familiar with the XP line so I can't tell you how to check (I only know the EVA line).
Another thing you could check is your san switches, check out how much IO is being sent via the different paths.