The trend towards locating data centre facilities near cheap sources of power has taken on new life with the recent release of two studies on the distribution of compute loads across data centre networks to take advantage of energy price differentials. A Canadian research initiative, conducted simultaneously under the umbrella of
CANARIE’s Green IT Pilot Program (the Canadian Advanced Network and Research for Industry and Education) has employed a similar technical approach to resolving issues surrounding data centre power, but to a different end. The overall goal of the CANARIE-led project is to develop a “zero carbon Internet” – one that relies not only on cheap power, but on totally green sources of energy.
The US reports, “
Cutting the Electric Bill for Internet-Scale Systems” authored by researchers at MIT, Carnegie Mellon University, and networking company Akamai, and the Rutgers/Princeton report, “
Cost- and Energy-Aware Load Distribution Across Data Centers” have commanded considerable media attention over the past week. In the MIT/CMU/Akamai report, researchers argued that software (provided by firms such as San Jose-based
Cassatt) now enables data centre service providers to direct information requests to facilities that have access to the cheapest power – the point being as much as possible to leave servers idle in network areas where and when energy costs are high. Through sophisticated simulation which factored in account spot energy prices at 29 US locations, load information from the networking firm Akamai, and a data centre electricity usage model provided by Google, the researchers were able to conclude that, depending on bandwidth costs, distance a request would travel, and the data centre’s own efficiencies, energy cost savings ranging from 2% to 30% (and up to 55% assuming an optimal dynamic system) are achievable. While this doesn’t replace the 65% savings that can be achieved by moving an entire facility to cheaper power sources, smart routing has the potential to offer an important economic advantage to cloud computing providers in particular who may realize significant savings by adding energy cost policies to systems that are already in place for ensuring fast delivery and bandwidth on demand.
