Here's something that I haven't thought about all the much, but is bound to become a more important consideration over time: If you want to outsource your data center to a third party to save some power, especially if that somebody is a services company that has both a product and services arm, be prepared to be told what equipment you can use.
I started thinking about this during a recent chat with Ed Kettler, a Hewlett-Packard fellow and green IT strategist with HP Enterprise Services, about their ongoing data center power optimization work -- especially the recent work the company's team has done in Wynyard in the United Kingdom, and Tulsa, Oklahoma. (Here's my original post on that work, plus a more recent one announcing the opening of Wynyard, which started taking on clients in February.)
Kettler noted that the racks being used in Wynyard are a key part of what is helping this data center run at a power usage effectiveness (PUE) rating of 1.2, which is actually better than was designed. Because Wynyard is a multi-purpose data center (meaning it is not focused on running one type of application or software solution like, say, a Google data center that is designed entirely for search), the rack design is especially important for HP services to control. There are data center halls in the facility, but they are handled separately.
One key design element of the Wynyard data center -- one that will be applied in other facilities as HP continues to look for ways to improve energy efficiency -- is the use of enclosed hot and cold aisles. HP takes great pains to keep the hot waste air from mixing with the cold, which means you don't have to work as hard to keep the cold air cold. Since I am a very visual person, I appreciated this video that HP has created to show this design principle in action. Sadly, they have disabled direct embedded of the video.
Anyway, all this means you'll need to confirm to HP's rack design, which will definitely be a consideration if you've got equipment that isn't designed for this.
Kettler and I also spent some time chatting about his concept of "Megawatts to Business Value," which is a cool way (no pun intended) for you to think about the value of different pieces of equipment in your data center. Kettler's theory is all based on the theory of "power bleed off," which is the amount of power that actually gets applied to a useful business purpose as opposed to the power that is applied to creating some sort of value for your company. This is especially important during data center consolidation projects. The fact it, Kettler suggests, sometimes certain legacy apps might be more expensive to keep running than pulling the plug. "Are you supporting a strategic investment or just keeping the lights on?" he asks.
Seriously, when should you pull the plug on an app that you may have inherited during an acquisition? Power considerations might help make that decision moving forward.
You know what, this one deserves an illustration, so I'm going to include HP's slide on this (even though it might be hard to read), which you can blow up and read at your leisure:
Although every data center scenario is different, Kettler says you can expect to see HP apply some of the aisle design principles from it Wynyard and Tulsa projects where it makes sense across its remaining data center portfolio.