With that in mind, Pouchet recommends that businesses establish an "energy policy." He likes to think of the data center the way he thinks about power consumption in his own house. "Everything is running in our houses because we might need it," he said. "We need to look at the data center the same way. Why is everything running if we don't need it?"
According to Pouchet, a typical data center has a lot of older, inefficient servers and a lot of "ghost" servers that no one knows what they're good for anymore -- but everyone is afraid to get rid of them. Does this sound familiar?
Pouchet's advice for businesses with "legacy" servers? "Hunt them down and kill them." (He did recommend migrating the applications off first.)
Among his other tips are ideas that many smaller businesses can implement:
- Low power processors (when you're ready to buy new equipment)
- High efficiency power supplies
- Server power management
- Blade servers
- Server virtualization
- Power distribution architecture
- Implementing cooling best features
- Variable capacity cooling
- High density supplemental cooling
- Monitoring optimization
It really all comes down to power, space, and cooling. Just improving the air distribution in a data center could make a difference, said Pouchet. He acknowledged that some of these strategies are expensive to implement but he believes that many pay for themselves within 18 months, at most. Some involve a "go forward" model ï¿¼ i.e. for servers reaching their fifth year, the policy is that new servers should meet the low power criteria. (Pouchet also noted that every watt saved at the server component level has a cascading effect and ultimately translates into 2.84 watts saved.)
By the way, Pouchet said that the last strategy he mentioned, monitoring optimization, should be a business' first move. He said: "You need to measure how much total power goes into your data center. You need to collect that data. How do you really know what's helping if you're not taking a look at your power bill?"