My PhD is about the use of on-site renewable energy sources (such as photovoltaic panels or wind turbines) in medium-scale data centers. Most of the renewable sources are intermittent and their output power vary a lot, depending on external factors such as the weather. However, a data center power need is usually not correlated to such weather variations. It is therefore rather difficult to count on those renewable energy sources alone for powering the whole data center. The electrical infrastructure may be improved by using energy storage devices (such as batteries, flywheels or fuel cells). Unfortunately, this does not fully solve the problem, unless one can afford ridiculous over-sizing of both storage devices and power sources.
Several complementary approaches are studied to use renewable energy more efficiently. In my work, two main ideas are explored:
- use the energy when it is available and reduce as much as possible the power consumption when required
- leverage the joint knowledge of IT and electrical systems to use the energy storage as efficiently as possible
In my recent works, I study how the IT and electrical systems of such datacenter can cooperate to schedule the energy usage in the near future. A negotiation module is used to find ideal compromise, satisfying as much as possible the objectives of each system.
- Grange, Léo, Georges Da Costa, and Patricia Stolf. “Green IT Scheduling for Data Center Powered with Renewable Energy.” Manuscript Submitted for Publication, 2017.
- Da Costa, Georges, Léo Grange, and Inès De Courchelle. “Modeling and Generating Large-Scale Google-like Workload.” In Proceedings of the Senventh International Green and Sustainable Computing Conference, First International Workshop on Resilience and/or Energy-Aware Techniques for High-Performance Computing, Forthcoming. IEEE, 2016.