Case Study: Virtualization Delivers A Cost-Saving Lesson

Antonowicz says it was a huge change in thinking when Bowdoin decided to go with virtual servers, basing so much of its network on a technology that was completely new to everyone there.

"It was just so tight in here. What a space crunch," says Antonowicz. "We were very tight as far as rackspace and resources so that we were not in a position where we could roll out new servers without impacting our cooling system, power systems, or KVM. We were running out of floor space."

He says another reason the college decided to move toward virtualization is that the IT staff was encountering problems with software vendors that were reluctant to support their products when they were running on hardware with other apps at the same time. "Cohabitation was not an option from a server standpoint," says Antonowicz. "We looked to virtualize our hardware infrastructure to create several virtual servers that would be standalone installations on a physical node."

Antonowicz started looking at Virtual Iron Software Inc.'s open-source Xen, Microsoft's Virtual Server, SWsoft Inc.'s Virtuozzo, and VMware Inc.'s ESX Server. "We went with ESX from VMware because it allowed us to virtualize the hardware for the Intel platform," explains Antonowicz, who says other products didn't support as wide a range of operating systems or management tools. "On a single ESX Server, we could have five virtual servers--one would be Windows 2000, one would be Windows 2003, one Windows XP, one Linux, and one Sun Solaris--all operating on the same physical piece of hardware."

Finding the right virtual software still didn't make the move any less nerve-wracking.

Antonowicz wondered if VMware would be stable enough to use in production. "It was kind of a new idea for us," he says. "VMware had been around, but we hadn't had a lot of experience with the resilience of the server product in everyday production with stresses, and we didn't know how well it would be able to handle backups, data recovery, disaster recovery, fault tolerance. It was a major concern."

It was enough of a concern that it used to keep him up at night. But it doesn't anymore.

Bowdoin had been testing the virtual servers for a few months when it suffered a hardware failure it simply couldn't figure out. Antonowicz believed the fastest way to bring the system back up was by running virtual servers. Virtual software management tools enable IT managers to quickly and easily move a virtual machine running on one physical server to a different physical server without interrupting the virtual machine. The switch can be done without lost data or downtime.

"I thought, I'm running this VMware stuff. Let me see if I can get it to work with that," he says. "I created a virtual machine and pulled it back as if it was a physical machine. I tweaked the drivers, and it worked. They got back online with it. At that point, I called up the CIO, Mitch Davis, and said, 'We need to buy this. We need to get this up and running.' It was a big step."

Over the next three months, Antonowicz and his IT team replaced four or five Intel-based HP servers that needed upgrading. Performance didn't suffer, he says. Actually, performance improved as they got processor and memory increases as part of the project.

Today Bowdoin relies heavily on virtualization. A majority of its applications, ranging from every flavor of Windows to SQL, Apache Web server, admissions packages, and giant financial applications, run on a virtual basis. In the next year, the school plans to switch another 15 of its physical servers over to virtualized servers, increasing the total to more than 100 virtual servers.

And other changes are coming as soon as this summer. Right now, says Antonowicz, he generally runs seven virtual servers on a single physical server. That number will multiply in the next few months when the college starts buying HP blade servers with dual-core AMD processors. At that point, he'll be able to run as many as 20 to 25 virtual servers on a single blade.

And Antonowicz says if IT had had to come up with the cash for a data center overhaul, it wouldn't have been able to implement a new financial system, or the system would have been severely scaled down. The new educational training system wouldn't have been rolled out. Web-based services wouldn't have been expanded.

"Business would probably have continued to go on, but it would have been severely hampered because a lot of the new services we rolled out in the past few years would never have happened," he notes. "We've been able to make the services we provide better."

Continue to the sidebar:
Virtualization Can Be Great, But It's Not For Everyone