The use of waste heat is indeed a possible source of heating in offices and spaces co-located with the center and there are examples of this use in many locations. Because the waste heat from servers is thermodynamically of low quality, i.e. low temperature difference from the ambient air, it is hard to extract useful energy from it other than supplementing heating load by mixing.

While servers and data centers are heavy users of energy, they should not be categorized as energy hogs. Servers now deliver 10 times the performance per Watt than they did less than five years ago, power efficiencies are now exceeding 75% end to end within the center and cooling power has been cut in half due to optimized center design practices over the same period. The total IT energy budget in the US is estimated to be no more than 2% of the total energy used. I submit that for the work the servers and associated systems in the data centers perform, at the energy they consume they are energy sippers, not hogs.

To be sure, we can do even better, but it will take more than physical design of cooling and power systems. Servers consume energy when doing nothing because they are built from clocked logic and memory, spinning disk drives and systems which are not designed to be shut down without long latency to come back online. The server platform can be developed to better match an ideal ‘load line’ where energy used is proportional to work done. Also, large groups of servers can be managed to maintain high server utilization, the most efficient operating condition, while putting as many less used servers into low power and even deep sleep states. Intel Labs developed this technology and recently began licensing it to interested software solution vendors. In lab tests we observed 10 to 15% energy use reduction for selected real world workloads.

I absolutely agree with the second point and look forward to participating in the national and world debate which will break down a century of regulatory barriers to energy consumer empowerment. We must not throw out the baby with the bathwater however. In the developed world, most regulations preventing microgrid to microgrid energy sharing and limiting microgrid to utility connections have been based on prioritizing the utility QoS first and utility personnel safety next. We have to convince the regulators that microgrids can satisfy QoS and safety before a major shift can occur. Microgrids also have to present significant economic opportunities for the companies we expect to supply the critical ingredients.

Social media can supply relative metrics to energy users. This has been shown to encourage competitive energy reduction behavior but it is too early to tell if it will result in long term behavioral changes. I’m interested to see how this plays out with experiments underway by Google (Power Meter) and Microsoft (Hohm). An interesting twist on this for those of us who are cube dwellers, is to seek out and eliminate office energy waste. At Intel we have an internal web site dedicated to encouraging employee input to energy reduction. I hope other companies are doing similar programs.

Tomm Aldridge
director, Energy Systems Research
Intel Labs