DV07 – Optimizing storage for virtual desktops
(Tony Armstrong – VMware, Mike Slisinger & M. Vaughan Stewart (NetApp)
To be honest, this was one of the session I didn’t like. I would have expected this session to be about how to work with storage in a VDI environment, but it turned out to be a commercial talk for NetApp and what they can do.
NetApp presented the following products:
-
Raid-DP. Its a fault tolerant disk setup like Raid10, but using only the amount of disks of a Raid config.
-
Data De-Duplication. This looks like ESX’s transparent page sharing for memory but now on disk. When identical blocks of data would have been written to disk, NetApp writes the identical blocks only once. This of course can save some diskspace, maybe even a higher percentage then transparent page sharing does because there is a lot more identical data on disk.
-
FlexClone for fast storage provisioning. A very nice technique which can deliver a 1000 VMs in a minute.
All the options presented by NetApp seem very usable in a VDI environment. Definitely something to look at when deploying VDI in a very large scale environment. Also check out the other storage vendors, because I was told by a colleague that he had seen similar options with other vendors.
De-Duplication is a cool solution for backing up data indeed. I worked with the DataDomain appliances in the past which definitely did what you would expect.
EMC’s Avamar is also a cool solution, it’s an agent which de-dupes the data before it’s send over the network and received on the SAN.
As a fan of NetApp and VMware for sometime the De-Dupe etc is all good stuff. But I was also took some interest in the Pillar Storage Products which claim to be “application aware storage”; I went to their presentation at VMworld but wasn’t 100% sure how they really differed from the others in the storage arena (apart from the 20% less than NetApp cost claim)
Pillar’s Application allows the admin to prioritize LUNs by writing high priority LUNs close to the spindles of the disk where access is faster. Therefore, if you run applications with high I/O requirements you can give those LUNs the prime space on your disks. We have an Axiom 3 in production and are running an Oracle DB and Exchange 2007 as VMs on high priority LUNs. I think this feature of their product stems from their relationship to Oracle’s CEO. You’ll need to look at your what you want to run to see if this feature has any value to you. If your interested in Pillar I will say this about them. Their service is very good and so far the product has been rock solid.