Store VMs on one or multiple SAN volumes

[TheTechie]

I like to get feedback on if we should store multiple VM OS drives on one or multiple SAN volumes in order to gain best performance.

We have a SAN and two servers to make into ESXi hosts. They will be hosting at least three servers each, two of the three will be setup in a two node MSCS SQL cluster so each hosts has one passive and one active node.

Our current plan is to put the clustered SQL drives on separate SAN volumes/LUNs using RDM so we can utilize the SANs snapshot capability. I also plan to use VMFS for the OS drives however is there any advantage to putting them on separate volumes.

For example have six separate SAN volumes, one for each VM OS or should we have one large volume per host setup with VMFS and store all three hosts OS drives within? Same question for the cluster Quorum volumes.

Thanks.

[Texiwill]

Hello,

I would have one LUN for the OS virtual disks and then use RDMs for the other disks needed by SQL.

Best regards,
Edward L. Haletky
VMware Communities User Moderator, VMware vExpert 2009, DABCC Analyst
====
Now Available on Rough-Cuts: 'VMware vSphere(TM) and Virtual Infrastructure Security: Securing ESX and the Virtual Environment'
Also available 'VMWare ESX Server in the Enterprise'
SearchVMware Pro|Blue Gears|Top Virtualization Security Links|Virtualization Security Round Table Podcast

[AndreTheGiant]

The VM OS disk can be on the same LUN of other VMs.

A good solution is have 2 LUN (on different disk pool) for VMs and separate the two node of the cluster.

Each RDM for the cluster shared disk must be different LUN.

Andrea

[savantsingh]

What i would do is to put the OS of your VMs on different VMFS volumes.(I don't think it enhances the performance really but you get peace of mind though!!)

For MSCS quorum disk why dont you use a shared RDM on a different pool.

[TheTechie]

Thanks for the feedback, I'll be taking this under consideration in the next few days as I try to plan this out.

"For MSCS quorum disk why dont you use a shared RDM on a different pool."

Are you implying just shared between the two nodes in the each cluster or something else, putting each cluster quorum (two) within the same RDM volumne?

[TheTechie]

This is a high level design I have based on feedback, probably have seperate LUN per databases but waiting to hear from DBA on requirements and depending on RAID requirements we may split them into another pool as well.

Storage Pool 1:

LUN1 VMFS: OS disk HostA (cluster1-node1, cluster2-node2, other VMs)

LUN2 RDM: Cluster1 Databases data

LUN3 RDM: Cluster1 Databases transaction logs

LUN4 RDM: Quorum for cluster1

Storage Pool 2:

LUN5 VMFS: OS disk HostB (cluster1-node2, cluster2-node1, other VMs)

LUN6 RDM: Cluster2 Databases data

LUN7 RDM: Cluster2 Databases transaction logs

LUN8 RDM: Quorum for cluster2

Would it be better to put the two Quorum in a 3rd storage pool?

We're using a 12x 750GB disk HP MSA2010i, would it be best to put each storage pool across all 12 drives or dedicate certain # of drives to each storage pool? might be a different thread post altogether.

Thanks again for the help.

[AndreTheGiant]

>Would it be better to put the two Quorum in a 3rd storage pool?

Is not neceessary, you can leave them on the same datastore.

Andrea

[habibalby]

Sent via BlackBerry® from Batelco

[savantsingh · Accepted solution]

Depends on how many cluster instances would you be installing..

RDM which is going to be used as a quorum should to be shared on both the cluster nodes(VMs).

I would not suggest to use the same RDM for a different cluster environment... you should use a different RDM for that..

This is just a suggestion there are many other things which you would need to think about from storage prospective ... if you give me more insight on your environment maybe i can help more..

[TheTechie]

Thanks for the help, not much I can add besides what in my previous posts but I think it solved. RDM for Quorum shared between cluster nodes but separate ones per cluster. VM OS drives can reside within same VMFS per host and put cluster instances into separate pools.