logo

That's great but what is the "N" in this equation? I have talked with several customers who are using the HA feature of VMware VI3 in order to provide N+1 HA. Typically, they are clustering together 4 ESX servers that are each running at 60% utilization (following VMware best practices). They are definitely seeing improvement in processing utilization as a result of using a hypervisor, sure. However, in this case they are operating in essentially a 2N high availability cluster, because 40% of the processing capacity is underutilized to allow for extra capacity in case of a hardware failure. Another way to think about this is that the "N" in this instance of "N+1" is only 3. I have seen this with other virtualization solutions as well, where the number can scale up to about 10. They reach their fundamental limit of underlying complexity based on the legacy server architecture. I've said it before - but it bears repeating - a new server architecture is needed to virtualize the I/O infrastructure that connects the server to the data center (eliminating physical NICs, HBAs and cables that connect the server to the network). If you can do this, you free up your host OS (and application) to move effortlessly between processing resources. In essence, your "N" can increase by a factor of 24! Before buying into N+1, figure out the physical and management complexities as the number of processing resources (N) scales. What's your "N"?