An Energy-aware Task Scheduler Based in Ownership Fairness Applied to Federated Grids

Abstract

The increasing complexity and the large volume of data used by applications lead to an ever-increasing use of high-performance distributed systems. The right management of energy consumption became a major challenge to hardware designers, application developers and administrators, since reducing power consumption and improving performance are conflicting goals. In this scenario, task scheduling policies can take into account energy consumption in their decision policies, helping to solve this problem. This paper presents the development and evaluation of a scheduling algorithm for independent tasks in federated computing grids. Its goal is to address energy consumption by associating it with an ownership fairness criterion. This is done by encouraging users to reduce energy consumption, or to provide energy-efficient nodes, by rewarding them to achieve fairness. Results from the simulation of this algorithm are presented.