It starts innocently enough: you start running virtual machines (VMs) for testing. Then you slowly get hooked. The next thing you start to do is put your VMs into production. Then you start buying only servers that have hypervisor built in. Then you start looking at SANs to leverage virtual clusters and virtualize your volumes. The next thing you know, you’re looking on vmware.com for a new virtual dog and banging your head on the keyboard when you can’t find one. At this point there are 8 VMs running on every laptop and handheld in your organization, there are sometimes 3 different applications running VMs on one host and often there is not a single backup of any of them.
You see, there’s a lot of upside with going virtual, but once you start down that path it’s hard to turn back, and nor would you want to for the most part. Buy many organizations make the plan to virtualize without factoring the full cost of doing so. This is because as a virtualization initiative begins to gather steam you start to realize some pain points that everyone encounters along the way. These tend to include virtual sprawl, lack of backups, the single point of failure on Host Operating Systems, etc.
So what do you do? Well, take a step back and think about what you’re doing. Put a plan in place. Consider storage as a good starting point as it’s often the single largest capital investment you’ll make on any virtualization project. Consider performance, both for the hardware that will run the VMs and the storage that they will sit on. Consider throughput, IO and what is best left unvirtualized. Then bust out your Visio or OmniGraffle and make a really nice network topology diagram that lays out where all of your pain points are. Then take that diagram and the costs you’ve figured out in your virtualization plan and find someone to help you make sure it makes sound financial sense. Then go to it!
But the problem we see is that too often the peer review of the plan by those in other parts of the organization goes undone. And this is critical. Because this will lead to buy-in far beyond your pay grade. Sure it’s fine to install VMware Server, Xen, XenSource or Parallels Server and run a 1 to 1 deployment of VMs in your environment but if you want to leverage all the bells and whistles that you’ll start needing to leverage, like ESX, clustered CPU and RAM, snapshots or SAN based VMs then make sure you understand the long-term implications in terms of cost and support before you go down that path.