Exchange 2007 double HA with VMWare and CCR?

During an internal discussion we came upon a very interesting doc/internal case study of Exchange 2007 deployment in VMWare:

One HA/DRS cluster is deployed in both of the VMware datacenters. The CCR design used at VMware populates each HA/DRS clusters with eleven CCR cluster-nodes. In the event of an unscheduled ESX host outage, the CCR cluster will automatically move the clustered mailbox role to the passive node in the second datacenter and continue to provide email services. Due to the VMware HA cluster, the failed active clusternode will be restarted on any remaining available ESX host, and CCR will initiate reverse replication to ensure that the cluster is in sync. The same methodology would be used in the case of a total site failure. To achieve complete autonomy between datacenters, a third site was chosen to house the file-share witnesses used by the CCR clusters to maintain quorum.

The paper provides a very interesting highly available design of Exchange 2007 with VMWare Cluster+ Exchange CCR. It would be tempting to go this way, as it allows to achieve a fairly high availability and site resilience w/o the need for expensive SAN metro-site mirroring.

However according to MS support policy this is unsupported (

Microsoft does not support combining Exchange clustering solutions (namely, cluster continuous replication (CCR) and single copy clusters (SCC)) with hypervisor-based availability or migration solutions (for example, Hyper-V’s quick migration). Both CCR and SCC are supported in hardware virtualization environments provided that the virtualization environment does not employ clustered virtualization servers

Another thing that came up is that one of the reasons not to virtualize Exchange mailbox is of course performance: VMWare and Hyper-V still have the maximum of 4 vCPUs. Whats more due to the way VMWare does CPU virutalization the more vCPUs are assigned to a VM the bigger the processing overhead. This is due to the fact that the VM must wait until there will be 4 physical cores available. It takes of course longer to wait for four cores then 2 or 1. Moreover if you have on the same host not only 4 vCPU but also mix that with 1 or 2 vCPU machines you will see more penalty (the 1-2 vCPU VMs will have a shorter time to wait, so might “clog” processors cycles of the 4 vCPU VM). This is due to the scheduling of vCPUs in VMWare virtual SMP. Some of it is described here: so basically it would be better to consolidate¬†2 vCPU machines on 1 host, and for large Exchange deployments with CCR/SCR opt for physical mailbox servers. I couldn’t find any information on how this problem is approached in Hyper-V, I imagine the situation is similar, but any feedback is appreciated!


~ by alipka on January 30, 2009.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: