Want to get started developing your server virtualization strategy but afraid of the potential traps that come with any new project? Tom Bittman, a research fellow at Gartner Inc. in Stamford, Conn., spoke with Margie Semilof, news director at SearchWinIT.com, about Gartner's list of best practices for IT shops looking to get the most from server virtualization. Here are his six rules of thumb for developing a virtualization strategy:
For those who want rapid return on investment, install server virtualization software when rolling out new hardware. Do it as you refresh your servers. When that happens, the vast majority of customers see ROI in three months.
Virtualize the right applications. IT shops are running into problems by trying to virtualize applications with high I/O, such as databases. Users should focus on more typical computing-intensive applications, and it's best to virtualize applications that underutilize hardware. Don't go for the applications that are transaction intensive.
Combine rapid-growth virtual machines. If you have a virtual machine that is active and growing at a fast pace and causing you to upgrade your hardware, it's better to put it with other virtual machines that are growing and not with static virtual machines. Run rapid-growth virtual machines at 30% utilization on a server -- not 80%. In terms of balancing risk and ROI, eight is a good ballpark estimate for how many virtual machines to host on one server, but the rule is not hard and fast. For example, some Gartner clients have 30 virtual machines on a server while some have just one.
Beware of software licensing. This is a big issue that holds many users back. Software is never based on virtual capacity. It is often based on the physical capacity of a box. When you take software on a two-way server and put it on an eight-way server, it will cost you more. Microsoft is very unusual in its licensing policy. Last year it launched a per-instance versus per-processor licensing policy. This will help push the market forward.
Think about process change. This isn't about consolidation. It's how you deploy your resources, and what to do for disaster recovery. A lot of people are using virtual machines for disaster recovery. Your capacity planning will change. How you do chargeback will change. From Day One you need to have a management plan in place for all of your virtual resources. There is a potential for virtual machine sprawl, just like server sprawl.
Have a long-term strategy. In many ways, virtualization is leading edge, and it's what forward thinkers are doing to modernize their infrastructure. But it's also used by vendors as a Trojan horse. They think 'I'll get virtual machines out there and then you will have to use my management tools.' But users need to know what they are doing with those tools.