Server virtualilzation and IT infrastructure?

Nothing will pull the covers on your planning abilities like virtualization. If you don't plan adequately, it will come back to bite you. Virtualization can work wonders for many orgs in many ways, but it's not trivial. Good planning will see you through, though.

In the old days, it was buy a server, configure it, and deploy it. If you needed another one, you repeated that. Now with virtualization, you have the ability to spin up new servers without additional hardware purchases. But you'd better get the underlying hardware right. Capacity planning and future expansion must be considered prior to buying hardware. A virtual server host is often going to be a very beefy machine, and depending on the size/scope of your business, it can scale up into SAN technology and virtually 100% uptime. Failure to look ahead properly paints you into a corner when you realize you need a few more VMs and don't have the hardware to support it.

The potential uses of virtualization are almost endless. My first rollout of VMWare at an ISV back in 99 resulted in a first-year savings of over $1M, as well as increased support satisfaction. For customer support, development, and sales teams, having a resettable sandbox environment not only makes their jobs easier, but saves IT time and resources; Support crashes the VM. Previously, it meant rebuilding and restoring their server. Now they do it themselves with a few mouse clicks.

User training on virtualization is almost unnecessary since it's so much like physical machines. But IT takes the hit on the planning, configuration, and deployment. Look carefully at your system requirements from a business POV. Expansion, performance, cost, redundancy, and uptime requirements should all be documented and signed off by management. They need to be aware of the final maximum capacity before writing the check. For significant deployments, the costs can be staggering, especially when you get into VMotion/SAN/HA stuff. There's also a need for your consultants and IT staff to be fully versed in the infrastucture technologies.

Management can be stunned by the initial costs, but the ROI calculations need to include things like recoverability; imagine being able to bring a crashed SQL box back online in an hour rather than a day; or disaster recovery; no longer do you worry about dissimilar hardware 2 years after a server is built; or malware recoverability; if your external webserver gets hacked a few commands and it's restored to a pristine condition. These are the points that overcome management reluctance, especially if you can cost-analyze your potential failure spots.

From an IT challenge perspective, once you've gotten the project approval, you must plan your operations and recovery strategies thoroughly, and test them regularly. Can you really move that VM to another host head? Is your snapshot recoverable? If you haven't tested in the past few months, you don't know.

Monitoring and tuning are, frankly, not much different from physical servers, although you do have some underlying aggregate counters that need to be looked at.

Planning for performance is also critical. You can throw a lot of processor/RAM at a VM host, but disk I/O is where your skills come in. Do you put a single VM on a spindle? Run a set of RAID 1 arrays? RAID 5? RAID 10? What is going to give you the best performance for each VM you spin up? A high-load SQL server is going to need very different I/O than a spoke-site DC.

Plan, Plan, Plan. Talk to the hardware vendor; see what support they can offer. Talk to the VM vendor. Talk to other businesses that have done it already. Talk to consultants who know virtualization. Before you spend a penny on hardware, you should have a road map of where you are, where you want to be, how you get there, and where you can go within the span of the hardware lifecycle.