How Efficient is Your Data Center?
I read a recent article from All Things D in which Arik Hesseldahl described a revelation he had during a lunch meeting with former Intel CIO (and current manger of Intel’s Data Center and Connected Systems group) Diane Bryant.
In the article, Hesseldahl describes Intel’s recently announced Romley generation of its Xeon processors. The Romley chips are both 80 percent faster and 50 percent more energy efficient that the previous generation Nehalem chips. Bryant was in New York promoting these new chips prior to her lunch with Hesseldahl. During the lunch, Hesseldahl cites a J.P. Morgan survey of one hundred CIOs in large companies showing that 91 percent did not expect to upgrade their data centers with the Romley chips.
Bryant responds by pointing out that she recently visited a customer who is one of the world’s largest 100 companies. At that particular customer, she notes, 36 percent of the chips in its data center were more than four years (or two chip generations) old. But that group of chips was consuming 65 percent of the data center’s power, while providing only four percent of the computing power.
Bryant’s counterpoint to Hesseldahl’s argument that the Romley chips might not be the game changer CIOs are looking for to upgrade their data centers shows that a simple calculation can reveal extraordinary inefficiencies in data centers. In the case of Bryant’s example, replacing the older chips will not only boost the data center’s speed and efficiency, it will also reduce the amount of energy needed to cool the computer towers.
So think about your data center’s age, capability, and efficiency sort of like a dog’s age. A dog that is five years old is actually about 35 in “dog years.” Similarly, your four-year-old data crunchers are probably about 80 years old in “computer years.” You might consider boosting your computing power and efficiency, and reducing energy expenses and emissions, by making a data center upgrade.