Many countries have set aggressive renewable integration targets. Achieving these targets requires fundamental changes to the management of the electric power grid. Data centers are considered as a significant consumer of electric energy worldwide. Furthermore, data centers are expected to grow their energy consumption footprint in the years to come and more data centers are expected to come online in the future [1]. Data center operators are acquiring power produced from wind or other renewable resources [2, 3]
Wind farms often generate power in excess of what is needed at a given moment in time and, for a variety of reasons, cannot be fed into the grid at time of production. It is conceivable that high-performance computing tasks, often scheduled as batch jobs, are ideally suited for computing scenarios that scale the computation up and down with the availability of energy directly harvested from renewables, Unfortunately, many enterprise and cloud workloads lack this scheduling flexibility. If cloud and data center loads could be scheduled on demand, the excess energy generated might be put to good use. However, the applicability of this idea to interactive workloads (e.g., clouds that back mobile and smartphone services) is challenging, as these workloads cannot be seamlessly scaled up and down.
The goal of this thesis to to Investigate the benefits achievable from co-locating data centers and wind farms, especially, for cloud, enterprise and interactive workloads. In validating this goals, this thesis should pursue the following four tasks (1) assess the state of the art as reported in the recent literature on co-locating high-performance computing facilities and high-performance computing workloads with renewable resource pools, such as wind farms, (2) repeat the reported results, (3) evaluate the proposed techniques on cloud, enterprise and interactive workloads, and (4) develop and experiment with software-based mechanisms to demonstrate whether or not cloud, enterprise and interactive workloads based data centers could equally benefit from co-location with renewable resource pools.
| Description of the task | Prerequisites |
|
|
Acquiring skills
A new master level course on ”Energy Informatics” where students will acquire knowledge and skills relevant for this project, will be introduced spring 2017.
References
- America's Data Centers Consuming and Wasting Growing Amounts of Energy.» [Online] February 06, 2015. Pierre Delforge
- Wind-Powered Data Centers [Online] http://www.datacenterknowledge.com/wind-powered-data-cente
- Google buys even more wind power for Oklahoma data center. [Online] 17 September 2013. By Yevgeniy Sverdlik