Link to home
Start Free TrialLog in
Avatar of JohnGerhardt
JohnGerhardtFlag for Switzerland

asked on

Hyper V (Cluster) over SMB via Scale out File Cluster or CSV

Hi Experts,
 
Can somebody confirm to me the major advantages of using a Scale Out File Server with a Hyper V cluster over just using CSVs?

I understand that with regards to failover SOFS are active / active so there is "no" downtime of a failure, but how much slower is a failover using plain CSVs?

What are ther other reasons to go for SOFS or not!

We are hesitant to add SOFS just for the sake of it as it would be another level of complexity that I am not sure we need.

Thanks for your input.

Jaggie

PS  - we are talking about Server 2016 here
Avatar of Cliff Galiher
Cliff Galiher
Flag of United States of America image

A SOFS deployment still uses CSVs.  Don't confuse which layer the technologies operate at.

A SOFS deployment would replace an iSCSI/FC/SAS storage array.  If you already have one of those, you may not want the cost of purchasing a SAN.  But if this is for a *new* deployment, SOFS has the potential to be less expensive than a high-end replicating SAN, which is the target market.

SOFS/iSCSI/FC/SAS operate at storage layer, so to speak.  CSV on ReFS/NTFS is the filesystem layer, and is relatively agnostic to that lower storage layer decision.
Avatar of JohnGerhardt

ASKER

Thanks,

Let me try and rephrase this. I already have the SAN.

As I see it for Hyper V Clustering I have two options:
  • I create a LUN on the SAN that is presented to both machines in the Hyper V Cluster via ISCSI.
  • This LUN is then added as CSV on each host.
  • Create a VM via the Failover cluster with the config and disk on the CSV

  • I create a LUN on the SAN that is presented to 2 (or more, up to 8) servers.
  • This LUN is added as a CSV on each of these servers
  • A Folder under the CSV disk is created and this is then shared via SOFS on each of the servers
  • A VM is created on an HV Cluster and this config and VHD are pointed to the SOFS share via UNC


Does this sound correct?

Cheers,

Jaggie
If you already have a SAN, you usually would *not* do SOFS.  Also worth noting, in your "rephrase" it wasn't clear whether your HV Cluster and SOFS cluster would be separate.  If you intended in collocating those, you *cannot.*  Your SOFS nodes would not run the Hyper-V role *at all.*  So you'd have physical nodes just for SOFS. And separate physical nodes in a different cluster just for compute (aka Hyper-V.)  

When using SAS JBODs, or in 2016 using Storage Spaces Direct, this eliminates a single point of failure. But when connecting back to a SAN, that is still a single point of failure. And when using multiple replicating SANs, SOFS would be duplicating the functionality that the SANs already provide. Which is why you normally wouldn't deploy SOFS in those scenarios. It'd be rolled out when replacing a SAN or similar. But not on top of a SAN.
Hi,

Thanks for this - I understand that SOFS and the HV Cluster would need to be on different Fail Over Clusters.

Are there any advantages though with regards to failover in having an SOFS?

Cheers,

Jaggie
ASKER CERTIFIED SOLUTION
Avatar of Cliff Galiher
Cliff Galiher
Flag of United States of America image

Link to home
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
Start Free Trial
Ok perfect.

That makes sense

Thanks for your advice.

Jaggie
Cheers