Power Management for Computing Clusters

University Of California-Riverside, Riverside CA

Investigators

Abstract

Dr. Brett D. Fleisch, University of California, Riverside PROJECT ABSTRACT: This research examines architectures, hardware, software, and algorithmic techniques to conserve energy in a distributed system cluster. The research focuses on improvement for better energy conservation. The research encompasses 1) monitoring and analysis and 2) improvement for better energy conservation including local optimizations and cluster-wide optimizations. For 1), the research will quantify power consumption costs for various configurations of networked systems. The research will gather and observe traffic in cluster systems that affects energy consumption and performance. For 2), the research will consider both local optimizations and cluster-wide optimizations that will improve energy conservation. For local optimizations, there will be low level improvements in device management, base-level communication algorithms, system daemons, drivers and systems applications. For cluster-wide optimizations, the research will consider a global process manager (GPM) similar to a load balancer that will move applications between sites based on criteria that includes both performance and power savings. The expected outcome of the research could suggest model topologies, networking technology, and software modifications for energy efficient distributed systems of the future. Lastly, the research could lead to improved algorithms for conventional systems software to operate in a more power-efficient manner.