An interesting article in Government Computing News (GCN), Comatose servers draining data center efficiency, points out the fact that there is still a great deal of waste when it comes to unused server power in the data center. According to the article, a new study by IT consulting firm Anthesis Group working with Stanford University supports earlier findings from the Uptime Institute as well as much of the anecdotal evidence we’ve heard from some of our clients:
30 percent of physical servers are lying comatose in corporate data centers
Call them “zombie servers”, call them “comatose”, they are costing corporations billions in energy usage. As we pointed out a while ago the NRDC estimated that as many as a full one-third of organizational servers are drawing power but are no longer in use by the organization for any production. Making matters worse, a single watt of power drawn at the server is estimated to consume 2.2 watts total between the server itself, transmission losses, and cooling requirements. Finally, older servers (those most likely to be dormant) were inherently less energy efficient at idle. The end result is nearly 40% of all data center energy usage is simply waste.
In GCN, writer Kathleen Hickey says that researchers have come up with two solutions for unused computing systems when it comes to government data centers. The first is to use distributed computing to eliminate the need for separate computers for different tasks. The second is to turn servers on and off depending on capacity needs.
Of course, that depends on being able to locate the individual servers in the first place, and as we’ve said before, that is often the biggest challenge. I think that the most viable third option, which can work with either or both of the previous options, is to tag and track data center assets throughout their life-cycles.
RFID tracking is now being leveraged by many data center operations teams to manage assets and automate the previous manual process leading to increased accuracy and productivity. It takes 10 percent of the time to perform an inventory audit using RFID technology that it used to take with manual processes.