7 things that will sink virtualization
- — 21 October, 2008 10:51
Many IT professionals measure the success of a technology deployment by all the things they did (see "7 tips for succeeding with virtualization"), but some say a successful virtualization implementation often can be the result of things IT didn't do.
"Many factors play here, such as the skill set of one's team, budget allocation and anticipated timeline. If you don't do your homework ahead of time [with virtualization], you can pay the consequences longer term," says Nick Portolese, senior manager of data center operations at Nielsen Mobile. "I learned this quickly enough midstream so I could benefit from it, but I do wish I had someone to tell me about all of this at the outset."
Here, industry watchers and enterprise IT managers reveal what not to do when starting or expanding a virtualization initiative.
1. Don't commit a false start.
Jumping the gun and rushing a virtualization project across IT will serve only to bombard staff with more work than necessary and fewer results, Portolese says.
Many IT shops start tinkering with virtualization on a small scale, but larger deployments require extensive research on servers, workloads, applications and business demands to reveal where the technology will benefit the company most. Portolese's advice starts with this: Don't rush a rollout until you have done your homework and know your environment inside out.
"An essential piece of this is to understand and categorize the workloads and existing management of each of the applications you want to virtualize -- before you start," Portolese says. "This has been a learning curve for me and my team."
2. Don't delay management plans.
Putting a detailed management approach in place after a deployment will result in performance problems across the virtual environment and headaches for IT staff, industry watchers say.
"Moving forward to expand virtualization without implementing good management discipline -- in areas like capacity, performance, configuration or automation -- is a recipe for disaster," says Andi Mann, research director at Enterprise Management Associates (EMA). "Once a deployment gets over a certain size -- around 100 or more virtual machines, depending on workloads -- or once it gets out of test and development and moves to production, the typically unmanaged environment is just not going to work."