Bag-of-Task Scheduling on Power-Aware Clusters Using a DVFS-Based Mechanism

Abstract

Energy reduction is very important nowadays. A large percentage of the workload submitted to large-scale systems is bag-of-tasks (BoT) applications. Each BoT is a collection of independent tasks that do not communicate with each other. They are used in astronomy, Monte Carlo simulations, data mining, fractal calculations, image processing and massive searches. Due to their importance, BoT scheduling is extensively studied regarding performance. In this paper we view BoT scheduling from an energy efficiency perspective. In order to save energy, we apply a Dynamic Voltage/Frequency Scaling (DVFS) mechanism to a heterogeneous cluster environment where BoTs are submitted. A cluster environment is selected due to the fact that clusters are often used as underlying basic components in grids and clouds. In order for our simulation experiments to be more realistic regarding the workload applied in the system, we also consider high-priority tasks. Extensive simulation experiments show that by applying the proposed DVFS mechanism when BoTs are executed, we can achieve energy savings up to 13% without affecting the execution of high-priority tasks.