Most of the larger-scale SharePoint deployments that I have been involved with over the last 6 years has included some type of virtualization included in the architecture, either through SQL virtual clustering, or with one or more of the farm server roles assigned as a VM Web Front End / Query, Indexing or a hot-spare application server. Many companies today are rapidly deploying fewer physical servers to handle all the SharePoint tasks of many physical servers and assuming that they can spin up as many as they see fit and obtain close to the same results as a physical environment.

Depending upon your particular method for determining the correct hardware requirements, you may or may not have taken into consideration the 'gotcha' aspects of virtualizing your SharePoint farm design – one that gets very little attention: Performance Monitoring

Performance monitoring in a virtualized environment, either through VMWare or Hyper-V is not as straightforward as monitoring a non-virtualized environment. I have seen one environment where 3-4 VM servers deployed had more resources assigned in total than the physical host had available. (Similar to banks leveraging 30x what they had in deposits over the last decade.) The VM is unaware of the fact that its configuration and assigned resources are virtualized. VMWare ESX and Microsoft Hyper-V based virtual environments share the physical state of the parent server. If you were to rationalize all of the VM CPUs, you would likely have a skewed view of the real system CPU utilization available within the host VM server.

For example, let's say you are beginning to see performance degradation on your SharePoint environment, so you remote in to you WFE and look first at the task manager 'performance' tab. You may be thinking, the server is underutilized, what could be the problem?

Although looking at task manager 'performance' and Perfmon is the place to start troubleshooting, if things seem ok here, you may need to go to the host / parent server for a more comprehensive view. (Can anyone say Governance here?) Since the CPU utilization you would see on the app server is only a snapshot of the local VM, it is in no way related to VM's parent server. This server is sharing the physical processors with the other VM instances in the farm. (To clarify, if architected correctly, with an even VM distribution, Perfmon can give you fairly accurate counters to help you tune your MOSS environment through page file, memory or application pool configuration changes.)

Everything looks fine here, so where do I look? VMware and Hyper-V have their own performance counters that are the best source of troubleshooting. These counters are related to non-SharePoint entities. Hyper-V calls them 'Guest' / 'Hypervisor' and VMWare calls them 'VirtualMachine' / 'Host System'. In the follow-up blog, 'Troubleshooting Virtualized SharePoint Farms - Part II', which will be published soon, I will give some detailed information around troubleshooting through performance monitoring. Until then, I hope that these links can help you with further information related to MOSS and VM performance troubleshooting:

Microsoft Hyper-V:

• http://technet.microsoft.com/en-us/library/cc816955.aspx
• http://blogs.msdn.com/tvoellm/archive/2008/05/12/hyper-v-performance-counters-part-four-of-many-hyper-v-hypervisor-virtual-processor-and-hyper-v-hypervisor-root-virtual-processor-counter-set.aspx

VMware:

• http://www.vmware.com/files/pdf/technote_PerformanceCounters.pdf