At Kelly Services Inc., a temporary employment placement firm located in Troy, Mich., the decision to virtualize was simple. The choices were to spend almost $30 million on a new data center or consolidate its servers. The company chose the latter.
"We were bursting at the seams," said Bob Brachulis, director of operations and services for Kelly Services, with upwards of 850 servers stuffed into 6,000 square feet of raised data center floor space. "Our air conditioning was maxed out and the building was out of power capacity."
The firm embarked on a plan to economize its power and cooling as much as possible through virtualization. It identified 269 older 1U, 2U and 3U Dell servers that could be retired over the course of the year and planned to migrate them onto Dell 1955 blades running VMware ESX when possible. Brachulis and his staff also resolved not to buy new equipment.
Before a single system was virtualized, Kelly Services' system architects and engineers identified possible virtualization candidates and used VMware Capacity Planner to predict how to use virtualization to get the best results. They then used monitoring tools from BMC Software to gauge the virtual machine's memory and CPU usage, as well as its responsiveness over time. With multiple physical-to-virtual migrations under its belt, Brachulis is confident that the company knows how to handle future virtualization projects.
Kelly Services first embarked on its virtualization project at the end of 2007 and began its rollout in April 2008. Since then, the company virtualized 69 systems across 10 ESX hosts and is moving toward a goal of 12:1 consolidation ratio. It also retired 50 of its targeted 269 servers and has improved its AC and power situation.
In the next phase, the company plans to add more management and oversight functionality into both its physical and virtual environments. A long-time BMC customer, the firm decided to resuscitate some of the "shelfware" it had purchased from BMC and put it into production--specifically the Service Impact Manager and Event Manager tools.
This should save time in figuring out the effect of server outages, Brachulis said. Event Manager and Service Impact Manager give the staff "a pictorial of what's impacting what and all servers that have been affected," Brachulis added. That information will also be integrated into the service desk console so administrators can see and fix problems before a call comes in.