This sounds like a case for CapacityIQ.
There are various counters to determine the IOPS and you could create another report using the datasource "logical drive" to determine drive letters used and capacity.
Last I knew of CapacityIQ it only dealt with size, memory and CPU and not IO. I'll have to take a look as it has been quite awhile.
See if vFoglight can provide what you need... Either the full version (30 day demo available) or their free vFoglight QuickView product... You could also check out the Veeam Monitor product (either the full version, or free version) to see if those will do it...
Consider awarding points for "helpful" and/or "correct" answers.
1 person found this helpful
you can use powershell script to get LUN disks statistics
http://wirtualizacja.wordpress.com/2010/02/02/powershell-pobieranie-iops-i-latency-z-lun/ is in Polish but you shouldn't have problems with copy/paste
visit my blog
if my answer it's useful click some stars
That script doesn't actually work for me...(returns zeros for the data) but I'll play with it. Plus it was giving the numbers per LUN - which by itself is helpful for the entire LUN - but unhelpful in determing which of the VMs accounted for the largest percentage. You could have 30 VMs on a LUN and never hit an IOPs bottleneck, or you could have 28 and need more spindles but only because of 2 or 3 VMs with abnormally high IOPs average compared to the rest.
What I really want is something that breaks it down like:
Guest A - 50 writes, 20 reads
Guest B - 20 writes, 40 reads
Guest C - 30 writes, 22 reads
Guest D - 23 writes, 17 reads
I've been working on a script using the PowerCLI that does something similar; maybe this will help you out? I made a quick post on my blog and linked to the source from there:
UPDATE: Sorry, forgot to mention that the credentials the script asks for are a local account on the ESX/ESXi hosts, NOT for vCenter.
This looks promising at least for "instanteous" numbers without having to look at esxtop. Thanks!
Ideally it would be something long term. Something that can end up giving me a pie graph of the ratios of IOps being done against a particular datastore to find out who the big hitters are and whether they represent an abnormally large amount that would be better suited for faster disk (or slower)