Energy Efficiency Oriented Scheduling for Heterogeneous Cloud Systems

Abstract

Nowadays, how to improve energy efficiency has become a challenging problem in cloud computing. However, most existing efforts in improving the energy efficiency of a cloud system only focus on resource allocation at the system level like managing physical nodes or virtual machines. This paper tries to address the energy efficiency problem of a heterogeneous cloud system at the task scheduling level. A novel task scheduling algorithm is proposed to reduce the energy consumption of the system while maintaining its performance, without closing or consolidating any system resources such as virtual machines or storage systems. The algorithm dynamically monitors CPU and memory load information of participating nodes on a heterogeneous Hadoop platform with Ganglia, then selects and submits an appropriate task to the node with relatively low workload to avoid excessive energy consumption on some nodes. Experimental results show that this novel scheduling algorithm can effectively improve the energy-saving ratio of a heterogeneous cloud platform while maintaining a high system performance.

Dynamic Job Scheduling Strategy for Unreliable Nodes in a Volunteer Desktop Grid

“…Energy efficient scheduling algorithm is proposed namely 'EEScheduler' that monitors the CPU and memory load of the participating nodes dynamically. EEScheduler performs the job scheduling by selecting a lightly loaded node for job execution such that it eliminates the excessive energy consumption from overloaded nodes (Lin et al, 2014).…”