Well, VMM 2008 is finally here. The folks that have played with this product the most up through now have found it simple to use yet a little buggy in early betas. I imagine those bugs are all fixed up more or less.
If you don’t know what VMM 2008 is, it’s basically the management part of Microsoft’s Hyper-V strategy. Hyper-V is the base hypervisor (compare with ESX) and VMM is the management layer on top (compare with Virtual Infrastructure). VMM is better for managing lots of Hyper-V servers than creating your own MMC.
From a storage perspective, we’ve found that a SAN that enables some of the more advanced features of Hyper-V and allows you to get the best bang for your buck. If you have virtualized and consolidated servers, having a vast amount of storage islands (DAS farms) is the opposite direction you’d want to go.
A SAN gives you a place to centrally store, configure, provision, and manage your storage and get the advanced features (clustering, HA, guest hardware-based snapshots, P2V without moving data, VM migration without moving data).
Table 3 in Jose Bareto’s Blog makes the case for a SAN quite well, I’d argue.
Some quick Hyper-V tips:
- Dedicate a LUN per VHD to allow for cluster configurations (and predicable performance).
- Make sure your array can support some type of Quality of Service to maintain LUN performance against other competitive LUNs within the array.
- Fixed is better than dynamic for controllable performance, but FlexVHD, although unsupported, could be something interesting to look at.
- iSCSI direct to the guest is typically going to be the best way to configure Hyper-V in the near term. Bypasses anything that might be blocked when doing passthrough and provides a semi-solid link (iSCSI qualified name) to the storage. Hyper-V clusters can then failover with no reconfiguration needed.
- You need to use the Enterprise version of W2k8 x64 in order to properly use Hyper-V and clusters because the Standard version namely does not offer clustering.
Here’s even more Hyper-V best practices from EMC’s Celerra team which includes the use of thin provisioning, LUN expansion, iSCSI vs CIFS considerations, replication, and more.