The other day while out shopping for Energy Star rated household appliances with my wife, I got to wondering why we still do not have Energy Star rated servers. This really is not as hard as some would have us believe. Even though it is not an exact science – witness the legion of consumers buying appliances on looks and not total energy consumption – it is still the best program we have to ensure energy reductions regardless of use behavior.
It is time for the U.S. EPA and industry to put their differences aside, line up a representative sampling of available servers, say in 2-socket, 4-socket, and 8-socket categories and start testing. Users need real data on industry average power consumption. Talking about memory, disk drive types, and management layers cloud the issue. With accurate data performance, criteria can be established to define Energy Star. From there, new innovations will come forward that will drive total energy consumption down. Over time today’s innovations that reduce energy consumption will become the norm by which future Energy Star ratings are defined. The bottom line is we need to start somewhere.
Let’s stop the debate over server classification, configurations, spec test protocols, etc. and get going building a meaningful database of server power performance. This could be easily patterned after the excellent EPA refrigerator database, where one can sort by size, top/bottom freezer, ice-maker, percent better than standard, etc. Such a database of servers would enable IT professionals to source more efficient (assuming real performance is similar) servers simply by specifying Energy Star. That may not change use patterns, but at least we then know the servers are burning 25 percent less energy than the standard server.
Emerson Network Power’s Energy Logic model demonstrates that every Watt saved at the processor could translate into 2.84 Watts at the building entrance. So switching to Energy Star servers that consume 25 percent less across the board and 50 percent less at “idle” (on but doing no work) could result in huge data center level savings. Assuming 100 W savings at idle and 60 W at nominal loads, the data center savings could be around 200 W per server. A small center with 100 servers would potentially save 20,000 Watts per hour or 175,200,000 WHr a year.
Most of us are already well versed in using EPA tools such as MPG for automobiles and Energy Star for household appliances – including complicated systems such as air conditioners – so let’s get moving on a rating for servers. Contact your suppliers, the EPA, or your Congressmen to demand some action.
We at Emerson certainly are, and we are prepared to help in the process.
Jack Pouchet is director of energy initiatives for Emerson Network Power.