Microsoft and VMware blog wars, the Empire Strikes Back (Exchange 2010 DAG)
Update: Microsoft has changed it’s support stance, see the below article.
I agree that using the Exchange DAG functionality within 2010 is the best High Availability solution you can implement to protect your mailboxes and provide the best uptime available. VMware HA alone does not provide an alternative solution, therefore due to the MS support requirements, you must disable HA by setting the VMware Cluster HA restart priority to disabled for Mailbox VMs within a DAG.
The MS article bases it’s case on not using VMware HA for DAG Mailbox Server as it is a better application aware HA solution and the additional costs of VMware HA. Now as stated above I agree on DAG’s being a better solution but costs? Every version of vSphere, (except the free version) licenses unlimited number of VMs for HA. If a customer already has vSphere then there is no additional costs. The only other requirement for VMware HA that could be perceived to have a cost associated is shared storage.
The primary advantage of Shared storage is also lost when using DAS, host failure means the data stored on that host cannot be used until the host failure is resolved. Well you might say fine, my database copies will activate on another DAG member. True but the surviving DAG member will be required to run more databases until the failed DAG member is restored. Depending on the load, size of your hosts and number or users, it is more than likely that during this time the users will notice slower performance. Where as an Exchange DAG using shared storage, another host can bring those database copies back online quickly and the databases can be redistributed back across two DAG members.
Therefore I do see benefits in continuing to use shared storage and do not believe cost is a significant hurdle to using a DAG.
Shared storage can be anything from iSCSI from an Openfiler server, NFS or Fibre Channel. Storage is a major component when designing and deploying MS Exchange 2010 with or without DAG’s. Exchange 2010 single instance storage is gone, for each database copy you plan to have increases the amount of storage required, plus you must have a restore volume. Therefore it is common depending on the level of protection required to need up to two or three times the storage you would normally require without DAG’s. This is clearly why MS pushes JBOD and cheaper DAS, instead of RAID FIbre channel so that your Exchange 2010 project doesn’t break the budget. Generally though Shared storage prices have reduced considerably and a lot more companies now have a SAN or NAS with SAS and SATA for either their current physical or virtual environment. If you do have existing equipment DAS in fact can be more expensive as it must be managed separately, it’s no cheaper to purchase and may require additional or different backup technologies.
I think this is another Microsoft blogging blunder much like the old blog wars over memory oversubscription. It is also clearer than ever before that Microsoft simply do not understand storage and the changes that have occured across the IT industry driven by companies Virtualizing their workloads.