Early in my days as a Data Center manager I attended a series of talks focused on Data Center energy efficiency. The sessions covered everything from hardware chip design to application performance to physical infrastructure.
Even for a beginner, two things were immediately obvious. First, Data Centers consume more energy than other buildings – much more. Second, with so many different components drawing power there are a lot of opportunities to make a server environment more energy efficient.
One presenter, from a manufacturer of Data Center standby electrical systems, mentioned during his talk that electrical components operate more efficiently at higher loads. The closer they are to maximum capacity, the better they perform.
I thought about this for a while and at the conclusion of the session, asked: “If electrical systems operate more efficiently at higher loads, why do operators of Data Centers with redundant electrical infrastructure split the load evenly between the A and B sides? Why not put the entire load on side A and nothing on side B? Wouldn’t that be more energy efficient?”
To my surprise, the question stumped the presenter. Eventually, one of his co-workers in the audience stood up and said they had conducted experiments with that configuration and found that although it was more energy efficient, when a failure occurred on the A side and the full power load (in his words) “came crashing onto the B side,” the components sometimes failed. The redundant electrical infrastructure could reliably handle a sudden jump from 40 percent loaded to 80 percent, but not from zero to 80 percent.
Oh. Enter my third Data Center lesson for the day: energy efficiency is important, but ensuring availability is much more important.
Speaking of availability and Data Center power, this week’s question explores the use of rotary UPS systems that employ flywheel technology versus traditional battery UPS systems. See below for discussion of the pros and cons of each.