I am rather excited to see Pushan Rinnen and the good folks at Gartner tackle the subject of storage choices with VMware in their recent report “Choosing Network-Attached Storage for VMware: What You Should Know” (Gartner report ID G00159391 published July 16th, 2008).
I’d like to share with you a few highlights from this report:
“Gartner found that most adoption of NFS to support VMware infrastructure is driven by ease of management especially when dealing with a large number of VMs”
“The current dominance of FC SANs in a VMware environment is further influenced by the fact that early versions of VMware did not support iSCSI SANs or NFS”
“For those who have decided to adopt NFS for VMware, they should be aware that not all NAS products are created equal and need to carefully evaluate specific NAS products that can cater to their requirements”
The content of the report is spot on with what we have experienced with our customers. Note: there was a wealth of additional information in the report, but you’ll need to leverage your Gartner and check it out for yourself.
Now none of this information should be earth shattering to you
Very similar information to what was contained within this report has been shared in recent months on various blog posts, including endorsements of running VMware with NAS storage from others within the virtualization and storage industries:
Joint Testing by NetApp & VMware
Recently NetApp & VMware have published the results obtained from a set of tests where the goal was to compare the relative performance of FC, iSCSI, and NFS with ESX in a production worthy architectural design (Technical Report TR3697).
The test bed included multiple ESX nodes, multiple Virtual Machines, concurrent datastore access by multiple Virtual Machines, and formatted file systems within each VM. In addition, the I/O load was generated from within each VM.
The results of this report show that all three storage protocols perform within 7% of each other under the heaviest workload ran in the testing (which trust me was way higher than I’ve ever seen a production ESX host run at).
This is great news for every VMware and NetApp customers as it validates the viability of all three storage protocols for production usage in enterprise deployments. In addition, these results support the notion that they are not locked into a storage architecture anymore than they are locked into a particular server design.
So what does all of this data mean to you?
If you prefer to run VMware on Fibre Channel or are already deployed on iSCSI, there’s no need to run out and change your infrastructure (unless you’re not on NetApp); however, if you are in the middle of considering deploying over NFS then I think you should feel confident that the world is taking notice of this design. As I always say, virtualization changes everything.