Let's say I have 4 servers in a "storage cluster", each of them has access to a JBOD of 24 hard drives. Then I have another 16 servers in a "VM cluster", running Hyper-V, each of them only has a boot disk. All of the servers are connected to a 10GbE switch.

S2D is set up on the "storage cluster", and the rest of the 16 servers have their VMs residing on the network share from the "storage cluster".

So I'm reading lots of information around various blogs, but I still can't figure it out: is there any real benefit of setting up SOFS on top of S2D?

  • 1
    In order to answer the question, you should check the high availability and reliability options you get from S2D. 4sysops.com/archives/… – A.Newgate Sep 27 '18 at 16:24
  • 1
    I'm aware (I think) of S2D's capabilities on it's own. What I'm not exactly sure about is what do I gain and what do I lose, if I set up SOFS on top of S2D... – Brane Sep 27 '18 at 17:20

If you don't set up SOFS, your file server (cluster) has a single point of failure. If one of the storage nodes dies, all the shares that it currently owns will fail over to another node, but all clients will be disconnected, which means that all of those VMs will crash. The share will come back, and the VMs will eventually restart, but this will be very disruptive.

With SOFS, a failure of a single node in the storage cluster might cause a slight stutter in the rate of I/O completion, but all requests made to the file server cluster will complete and the VMs will continue to run.

| improve this answer | |
  • Ah, that makes more sense now. A quick follow up question: can the SOFS role be installed on different machines than S2D, or should it always be installed on machines that are members of the S2D cluster in order to provide protection against unwanted disconnections? – Brane Oct 10 '18 at 11:57
  • 1
    Scale-Out File Server Role should co-exist with S2D. BTW, you should be aware of the fact SoFS isn’t a clustered file server replacement, it has no cache so performance for non-cached by default workloads (Hyper-V, SQL Server etc) is simply horrible. docs.microsoft.com/en-us/windows-server/failover-clustering/… – BaronSamedi1958 Oct 13 '18 at 12:43

S2D is uber-expensive for what it does in terms of storage-only deployment. It's Datacenter-only WS feature so while HCI scenario kind of makes sense - segregated / SOFS definitely doesn't. I'd leave things AS IS. If you absolutely want SOFS for whatever reason - look @ StarWind VSAN Free. It's much faster, more reliable within 2-Node deployment and... it's free :)

| improve this answer | |
  • 2
    Microsoft had actually released an update facing issues you're referencing. Google for "nested resiliency" which is nothing but RAID51 both StarWind and HPE SimpliVity have been doing for years now. – BaronSamedi1958 Sep 30 '18 at 10:46

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.