The energy cost of operating large server farms is now a sizable portion of their total-cost-of-ownership. The focus of the ecoDB project is to design, develop and evaluate methods that improve the energy efficiency of such server farms for data-intensive applications. ecoDB will investigate a range of issues, including "global" issues that considers the entire server farm as a holistic single large computing system and uses energy-aware workload management and data placement strategies. These global techniques will be implemented in various existing distributed systems, including the Condor system. At the other end of the spectrum, this proposal plans to investigate "local" techniques that can be used to improve the energy efficiency of an individual server by synergistically exploiting the underlying hardware and/or software characteristics. A crucial aspect of the ecoDB project is to focus on techniques that systematically trade performance for energy efficiency, essentially treating "energy consumption" as a first-class metric in data processing systems. The repercussions of this approach percolate through various aspects of a data processing systems ranging from query/workload optimization and evaluation to replication management and job scheduling in large-scale parallel and distributed data processing systems.
This project will also facilitate the training of graduate students in the emerging area of energy-efficient data processing methods. The broader impacts of this proposal also include benefits to society by producing techniques that can potentially reduce the energy consumption of data centers, which in turn has beneficial environmental and economical effects.
This project is partially funded by the OCI CF21 Venture Fund for promoting the reuse of Cyberinfrastructure (CI) elements.
For further information, please see: http://pages.cs.wisc.edu/~jignesh/ecodb/
