Windows Server 2008 R2 Failover Clustering, which is a standard component of Hyper-V Server 2008 R2, is changing the support statement for the maximum number of Virtual Machines (VM’s) that can be hosted on a failover cluster. Originally it supported a maximum of 64 VM’s per node. As of June 8th 2010 we changed this to 1,000 VM’s per cluster. So the limit is no longer bound to the node but to the cluster as a whole!
Compared to the original support statement we now actually support more VM’s per cluster. You may wonder how I calculate that, as we originally supported 1024 VM’s per cluster…..
Running 1024 VM’s on a 16-node cluster was more of a theoretical maximum than a real world maximum. Because running a 16-node cluster with 64 VM’s per node leaves no room for failover. A 16-node cluster should have room for failure of one node and for maintenance. This would mean that you would calculate only 14 hosts with 64 VM’s each which results in 896 VM’s in the cluster. With the new support statement you can actually run 1000 VM’s in the cluster which calculates to 71 VM’s per node in the above scenario (rounded down).
You can now configure an 8-node cluster with 166 VM’s per node, considering two hosts for failure and maintenance. 166 VM’s may still be theoretical as most hosts would not support the required disk and network throughput. But when most of the VM’s are idle, this scenario will now be officially supported. You may wonder about failover in such scenario’s but failover of such high densities (and even higher) has been tested and proven to work well.
So the new support statement helps achieve higher VM density. It enables increased flexibility to utilize hardware that has the capacity to host more VM’s per physical server. Simply put, you can run more VM’s in your cluster.
Notice! It is still important to perform proper capacity planning that takes into consideration the capabilities of the hardware and storage to host VM’s and the total resources that the individual VM’s require, while still having enough reserve capacity to host VM’s in the event of a node failure to prevent over commit. Any base guidance of Hyper-V configuration and limits of maximum number of VMs supported per physical server still apply. Which is currently that no node can host more than 384 running VMs at any given time, and that the hardware scalability should not exceed 4 virtual processors per VM and no more than 8 virtual processors per logical processor.