Hey All,
I'm trying to get a handle on IOPS figures for the workstations I'm planning to move to View 4, using Windows perfmon to collect the data. While the averages are within the range I expected to see (4-7 IOPS), I'm also seeing spikes that go well over a hundred IOPS and a fair number of them too. I don't quite know what to make of these.
Based on the average IOPS I have plenty of throughput available on my SAN to handle the load. Any advice?
I would consider working with your storage vendor on this. I took a perfmon for 30 days on our test machines in VDI - having one of our "power users" Pound on it for a month. I then Took the average of the total for the day and times it by amount of desktops to be safe. You have to really work on the probability of the desktop hitting their max I/O. You will also have to take into consideration if you are going to use composer or not. Composer will save you on some I/O.
I would consider working with your storage vendor on this. I took a perfmon for 30 days on our test machines in VDI - having one of our "power users" Pound on it for a month. I then Took the average of the total for the day and times it by amount of desktops to be safe. You have to really work on the probability of the desktop hitting their max I/O. You will also have to take into consideration if you are going to use composer or not. Composer will save you on some I/O.
Perhaps the info on this blog post will be useful?
http://www.vmguru.nl/wordpress/2009/12/vmware-view-sizing-best-practices/
I'd check out this blog post from Herco , he has done extensive investigation in IOPS.
greetings
Rob
Thanks guys, I'll take a look at those items.
Agreed, thanks for those links
EMC has just pushed a new white paper that talks about sizing in general and has some good suggestions on IOPS estimation. The paper is here.
I found this through Reuben Spruijt's tweets. He's also on BrianMadden and has a recent blog here.
- Eric