Doing the math on virtualization

Network World - Money is tight. Performance is declining. Your servers are all nearly three years old, and pretty soon, their high-priced maintenance contract is about to kick in. What do you do?

If you're like Tim Hays, you take the money you would have spent on maintenance and instead use virtualization to not only improve performance, but also increase IT efficiency, cut power and cooling costs by 45%, and make disaster recovery as easy as pushing a button.

"I guess I'm more of a business guy first and an IT guy second," says Hays, director of IT at Lextron, a wholesale distributor of animal health pharmaceuticals with 600 employees in 44 locations across 19 states. "I look at IT as a business enabler. Virtualization wasn't something that we did just because it was the next cool thing. It made a lot of economic sense."

Start small

All told, the three servers were responsible for handling $1.75 million in sales transactions daily.

At the time, virtualization a la VMware was not well known. Instead, he used the $300,000 to replace the Unix servers with one PA-RISC-based HP-UX server running HP's Virtual Partition (vPAR) software, which enabled the one server to host three virtually partitioned servers. He also put the server on a Fibre Channel-based HP EVA 5000 storage-area network. (Compare storage virtualization products.)

"Cost-wise, it was a wash," he says, noting that the depreciation for the new equipment was $100,000 per year, the same as he would have been paying for maintenance per year on the old gear.

But reports now ran 30% to 40% faster, and user complaints declined. "People were waiting less time to get information, and they didn't have that problem where they were outworking the ability of the system to retrieve the data," he says.

At the same time, the company also needed to upgrade its ERP databases from Informix 7 to 9.4. "We had 1,500 programs that we needed to regression test against new hardware, a new database, new development tools and a new operating system," Hays says. Buying new equipment enabled Hays to install all the new software and thoroughly test everything before cutting over. Once his team was confident everything would work, they simply switched users from the old equipment to the new. The whole process, which could have taken six months in the past, took just 45 days.