When Supermicro donated an energy-efficiency optimized server to the Shinsai portal project in the wake of the earthquake and tsunami disaster to hit Japan, they were doing a good deed to help a community in need as well as a company using their core competency to deliver a service that was needed in the aftermath of the disaster. The Shinsai portal site has become a central clearinghouse for news about charity, relief, and community building efforts for the victims of the disaster.
Now it would be easy to just pat Supermicro on the back and say "nice job." But I think it is important to consider what lesson can be learned from this. We tend to look at energy-efficiency as another checkmark on our "what I need in a server" list, but the situation that Japan found itself in merits another look at that checkmark. Consider this; every energy conserving methodology you employ in your datacenters converts directly into how long you will be able to stay up and running when you have to go to your backup power.
Most datacenters with significant disaster strategies in place start with backup power generation. The greater the load on those generators, the faster the fuel supplies will be consumed. And as I talked about a few weeks back, refueling may become impossible, just when you need it most. We calculate the load we need to support when specifying backup power generation, but as technologies change it becomes possible to increase the workload without changing the total power demand. In large datacenters we tend to take that as a given, as I replace older technology, newer technologies will be greener and reduce my overall power demand, so that number I used when the datacenter was built 5 years ago (for example) is more than adequate for my new infrastructure.
But not every datacenter is on a scale when that over-provisioning of backup power is possible, or likely. Many smaller datacenters are already in a position where loss of primary power means the potential shutdown of services if the power outage is at all protracted. For those living on the edge of their power demand in a disaster situation, the issue can become paramount. In remote datacenters or server rooms, the ability to run on backup battery power, for example, for an extra few hours may be the difference between having to keep significant backup resources on site or have other, more cost effective, ways of taking over the local workload.
What Supermicro's donation reminded me of was the need to consider just how powerful my servers need to be. If I can effectively service my computing needs while making use of lower power CPUs and slower, but more efficient servers and storage, without negatively impacting my users, I need to place a higher priority on evaluating just how much work can be done with those more power efficient, but less compute capable systems.