Server 2012 - Clustering CSV Config advice

Question

We have a Dell VHDX Server with 2 server blades (This has shared storage built in)
I've installed Server 2012 Datacentre on both servers, configured a cluster (My first!)
I have a number of HDD's in the chassis and hardware RAID is built in.
A file-server VM, print server as well as other miscellaneous VM's will be installed on the cluster.
What would be the ideal config for the CSV's? Should I create a RAID CSV exclusively for the VM's themselves to reside on?
Then separate RAID CSV's for each data requirement?
What about the SQL Server, should that have its own separate CSV? (Its not heavily utilised)
For the File-server, should the primary group shares be in the file-server VHDX, or a separate VHDX for each?
Should I consider using SMB 3.0? and how would I configure/design that?
Any advice or suggestions appreciated.

Answer 1

We have a Dell VHDX Server with 2 server blades (This has shared storage built in)
I've installed Server 2012 Datacentre on both servers, configured a cluster (My first!)
I have a number of HDD's in the chassis and hardware RAID is built in.
A file-server VM, print server as well as other miscellaneous VM's will be installed on the cluster.
What would be the ideal config for the CSV's? Should I create a RAID CSV exclusively for the VM's themselves to reside on?
Then separate RAID CSV's for each data requirement?
What about the SQL Server, should that have its own separate CSV? (Its not heavily utilised)
For the File-server, should the primary group shares be in the file-server VHDX, or a separate VHDX for each?
Should I consider using SMB 3.0? and how would I configure/design that?
Any advice or suggestions appreciated.
Looks like it's an extended version of a post from here:
http://social.technet.microsoft.com/Forums/windowsserver/en-US/2d7cd5b0-cbe5-4dcd-8ed1-759c12cc81ae/server-2012-csv-raid-advice
1) Don't go for multiple LUNs as that's an effective way to steal IOPS from your config. Put all spindles into RAID set and create one RAID10 LUN to host one CSV. Reasons are listed with the link above.
2) Don't create multiple "islands" of storage. Hell to manage and lower performance.
3) Don't separate moderately loaded SQL Server VM from other ones.
4) You map SAS disks AS IS, creat Clustered Storage Spaces, create a CSV, you create file server VMs and keep shared content on a VHDX stored on CSV. See:
http://technet.microsoft.com/en-us/library/cc732181(v=ws.10).aspx
With R2 you can use shared VHDX instead of iSCSI running on top of your storage.
5) With two blades you cannot use SMB 3.0 as you cannot run it in a loopback. See:
http://blogs.technet.com/b/josebda/archive/2012/11/14/windows-server-2012-file-server-tip-avoid-loopback-configurations-for-hyper-v-over-smb.aspx
With 4 blades you could in theory assign pair of blades to become Hyper-V hosts and another pair to become SoFS set but there's no point in doing this: you'll steal CPU resources from your hardware config (2 boxes instead of 4 running VMs) and make everything
slower (1 GbE interconnect built-in into VRTX). Use Clustered Storage Spaces as that was the idea behind whole VRTX thing.Hope
this helped :)
StarWind VSAN [Virtual SAN] clusters Hyper-V without SAS, Fibre Channel, SMB 3.0 or iSCSI, uses Ethernet to mirror internally mounted SATA disks between hosts.