UW News

November 9, 2006

A tale of three data centers

News and Information

In the world of computing, the Holy Grail is power.

In the old days at the UW, five or 10 years ago, central computing facilities would need to be expanded to house an ever-growing army of servers, the machines that store and process administrative information for the campus and also handle the critical needs of many research projects.

Now, more and more computing power can be crammed into fewer, ever more powerful computers — but the electrical and cooling needs of these devices continue to grow.

Computing & Communications, therefore, is continually searching for new and better data centers — centrally-managed facilities that provide restricted access, a controlled environment, abundant usable power with readily available backup power and impressive resources for cooling.

The good news is, the university’s short term needs for data centers seem under control.

Data centers already exist at 4545 15th Ave. NE and also at 3737 University Way NE. But both were running out of either cooling capacity (at 3737) or electrical power (4545). And acquiring more of these at either site was looking to be a pricey proposition. A committee chaired by Marilyn Cox, assistant vice provost for capital planning, examined the requirements for creating one primary data center and concluded that the complexities of siting the facility and acquiring power would make this approach unrealistic. Instead, a plan has now been approved to move to a model of centrally managed, but distributed data centers — keeping the 4545 Building data center, and a leased data center that was opened earlier this year, and developing new data center space in a third location.

The acquisition of Safeco Tower, now UW Tower, provides a convenient location for the new data center space in a university owned building.

As luck would have it, Safeco was built with the idea that one day it might be a data center, so it has most of the necessary infrastructure, especially power, much of which has not been used. “A data center at UW Tower has the basic reliability for many research and business systems,” says Scott Mah, assistant vice president for Computing & Communications. “It has space for growth, enough power and cooling to meet many of our needs for the next five to ten years.”

The UW has requested $25 million in capital funds from the state in the coming biennium to develop the data center at UW Tower. In addition, a research bond is being proposed to be paid over time by research indirect cost recovery funds to help with the cost of the facility.

Those computing resources that have been deemed “mission critical,” such as payroll and email, will continue to be housed at 4545. However, because that facility is so close to its maximum power usage, C&C plans to bring down the electrical load there through energy conservation projects unrelated to the servers — such mundane things as heating and lighting, which will provide a margin of safety on this heavily-taxed facility. In fact, it is likely that the building’s heating needs in the future will be met in large measure by “waste” heat from the building’s computers.

For other large projects requiring a high degree of reliability, the UW has rented space at a data center in Tukwila. The Tukwila facility, he says, actually promises a higher degree of reliability than 4545, albeit at a remote location.

A combination of these three data centers is likely to meet the university’s near term needs, but the issue of paying for the growing computing demands, especially for research projects, remains an item for discussion. The current proposal is that research projects needing central computing facilities will pay a portion of costs, perhaps a third of full and true costs. And the people at Planning and Budgeting and C&C are committed to exploring all options in keeping costs manageable.

“This is very much a moving target,” Mah says. “It’s impossible to project accurately what our needs might be in ten years.”

Nonetheless, C&C has begun investigating locations in eastern Washington, because some of the communities there, which own dams, have had power available under long-term agreements at reasonable cost. Mah points out that these locations are ideal for backup systems that would come into play in the event of a disaster, such as a major earthquake, in Seattle.