MetroCluster with VDI

[chriskranz]

Are there any considerations when deploying VDI on a MetroCluster?

Are there some recommended maximums for the number of clones, snapshots, and so on that should be taken into consideration?

Are there any guidelines for putting together the perfect datacentre with MetroCluster in mind? For example, should we look at a separate View Composer or Virtual Centre server for each node of the MetroCluster to allow for the best resilience?

[reinoud7]

I think it's a great platform for VDI, definitely when you combine it with dedup (you win twice).

Personally, I wouldn't take special measurements for Virtual Center and other stuff: just use the two controllers and the metrocluster take care of everything. That is just the strength of the system.

We use metrocluster today for our ESX environment: everything what you put there is automatic fault tolerance. It works great and I assume that it would the same with VDI.

Reinoud

[amiller_1]

To me the main thing here would be MetroCluster's lack of support for SATA disk (unless using a V-Series with something else behind it of course).

[chriskranz]

That limitation goes away with 7.3.x -

[amiller_1]

Fantastic -- am wondering if the FC --> SATA use case will ever become common enough to be supported.

[darrin · Accepted solution]

Q: Are there any considerations when deploying VDI on a MetroCluster?

A: No. All of that sits far above MC.

Q: Are there some recommended maximums for the number of clones, snapshots, and so on that should be taken into consideration?

A: No. MetroCluster places no restrictions in that area. The only potential factor is that as with an active/active configuration, as the total number of volumes go up, the failover time can increase.  And remember as you make vol additions, etc on either side the data will be mirrored automatically to the other side, no need for configuration or replication setup/etc..  It just works.

Q: Are there any guidelines for putting together the perfect datacenter with MetroCluster in mind? For example, should we look at a separate View Composer or Virtual Center server for each node of the MetroCluster to allow for the best resilience?

A: From a MetroCluster perspective, just follow the normal Metrocluster Guidelines outlined in the following report: http://www.netapp.com/us/library/technical-reports/tr-3548.html. From an overall availability and resilience perspective, there should be redundancy at the vCenter server level certainly. Ideally the vCenter server should be either redundant or reside at a third location (or both). Alone, the vCenter server is a SPOF and since set up of many functions (such as set up of vMotion) require vCenter, it should be resilient.

Thanks.

Darrin

[chriskranz]

Many thanks both Reinoud and Darrin. Wasn't sure if there was anything specific to take into consideration, so glad it stays nice and easy!