UPSes Keep Campuses Up and Running
Uninterruptible power supplies are low-profile workhorses in the data centers of higher education and other fields. That doesn't mean, however, that their role is any less vital to high availability than more glamorous technologies such as virtualization, storage area networks and server clusters.
Just ask IT officials at Fairfield University, who rely on their Liebert UPS to keep core systems in their data center, and a wide range of business and academic applications, humming. Since first deploying the UPS in January 2007, the Fairfield, Conn., university has come to trust the system through power surges, storms and other brief interruptions in power. That confidence has been rewarded with essentially flawless performance, says Steve Dailey, assistant director of computing and network services at Fairfield.
A UPS provides battery backup when electrical power to a system, or even a complete data center, fails or drops to an unacceptable voltage level. It delivers power temporarily to power down systems in an orderly fashion or to bridge the gap until backup power becomes available. Some versions can run systems for several hours if necessary.
At Fairfield, the IT team selected UPS technology after recognizing that many of the university's systems couldn't endure even brief interruptions in availability. “We wanted to protect from power spikes and give ourselves leeway in case we did have a power outage,” says Dailey. “Our staff has always taken the perspective that we need to get everything back up as soon as possible.”
That mindset led to the decision to invest in an 80kVa Liebert system, which currently supports about 100 servers in the university's primary data center. Those servers run university e-mail, administrative applications, databases, web services and more. The university has approximately 80 percent Windows-based servers, 15 percent Macintosh systems and 5 percent other platforms. Its network services travel across Cisco Systems switches and routers.
Maintain and Sustain
IT organizations need to understand the importance of individual applications and assign them priorities as part of a high-availability strategy, says Nik Simpson, senior analyst for computer and data center infrastructure at the Burton Group.
“You assess the applications, availability requirements and the downside of being unavailable for any length of time,” Simpson says. “There's no sense spending a lot of money for a high-availability power infrastructure if the apps it's supporting can be down for a couple days without anyone noticing.”
From 60 servers to 2 The virtualization target that Fairfield University has set as part of additional steps to increase availability
At Fairfield, the UPS is just one element in an overall power management strategy. The university powers its campus through a gas turbine that converts natural gas to electricity. There is also a backup generator in the event of a sustained power outage.
Experts say most organizations should invest in these types of long-term backup power delivery capabilities. There's no point in gaining an extra 15 minutes of runtime through a UPS alone if a power outage lasts a week, Simpson says. “You haven't bought yourself a lot of redundancy” in that scenario, he says.
Indeed, other universities with UPS systems and backup generators recognize that even their backup systems often require redundancy. For example, Louisiana State University is planning to increase availability in its primary data center by adding a second UPS and a second backup generator in the months to come, effectively putting all servers and applications deemed critical behind two UPSes and two backup generators, says Ric Simmons, deputy CIO and executive director of university networking and infrastructure. Those systems considered most critical at LSU include administrative computing applications, the campus' primary web servers and network connections supporting the university website, says Terry Doub, director of LSU's network operations and data center in Baton Rouge.
No Downtime Allowed
LSU's data center supports hundreds of servers and more than 100 virtual machines in its 14,000-square-foot facility. High availability is particularly critical, Doub says, because the facility also hosts servers for a handful of third parties, including state agencies and other universities – and in a state that can face extreme weather threats. “With all our resources in here, we want 99.999 percent uptime,” Doub says.
In Fairfield's case, success with its UPS has helped the university refine its approach to introducing new equipment at its data center. Now, any new devices are carefully scrutinized on the basis of their power requirements to determine optimal placement in the facility.
“We want to plan the best place to put it for airflow and to distribute the power draw, so we don't have one aisle of servers taking up 75 percent of the power,” Dailey says. The university recently experienced a system failure from a hot spot created in the data center.
Since Fairfield first installed its UPS in 2007, it has roughly doubled the number of supported systems. “The demand for high availability has increased, and the Liebert just keeps chugging along,” Dailey says.