¡¡	Chinese Journal of Computers   Full Text
Title	Energy Efficient Scheduling and Optimization for Parallel Tasks on Homogeneous Clusters
Authors	LI Xin1) JIA Zhi-Ping1) JU Lei1) ZHAO Yan-Heng1) ZONG Zi-Liang2)
Address	1)(School of Computer Science and Technology, Shandong University, Jinan 250101) 2)(Department of Computer Science, Texas State University, San Marcos TX 78666, USA)
Year	2012
Issue	No.3(591¡ª602)
Abstract &   Background	Abstract The design of energy-efficient scheduling algorithms has become a hot research topic in high performance computing. To shorten schedule length of parallel tasks with precedence constraints, scheduling algorithms could duplicate tasks on critical paths to avoid communication delay caused by inter-task dependence. However, task duplications incur more energy consumption. In this paper, we propose a heuristic Processor Reduction Optimizing (PRO) approach to reduce the number of processors used to run parallel tasks, thereby decreasing system energy consumption. The PRO approach can find appropriate time slots to accommodate tasks from low-utilized processors according to their earliest start time and earliest complete time. Extensive experimental results show that the proposed PRO approach, compared to existing duplication-based scheduling algorithms, such as Task Duplication Scheduling (TDS), Energy-Aware Duplication (EAD) and Performance-Energy Balanced Duplication (PEBD) algorithms, can effectively decrease the number of used processors and save energy without performance degradation. Keywords green network; cluster; parallel; homogenous; precedence constraint; energy-efficient scheduling; green computing Background This paper is a research product from our projects mainly supported by the National Natural Science Foundation of China (NSFC) under grant No.60903031 and 61070022, the U.S. National Science Foundation under Grant No.CNS-0915762 and CNS-1118043. These projects explore energy-efficient techniques or thermal-aware methods for scheduling parallel tasks in multi-core or cluster systems. We also develop simulator programs that can analyze the run-time power consumption of different system components (e.g. processors, network and disks). Now we are designing novel scheduling algorithms to achieve high performance, reliability and energy efficiency. Many research projects have taken measures to reduce power cost of data centers. For example, IBM Blue Gene¡k/L supercomputer makes use of low-frequency, low-power processors with modest performance to save energy. The other well known idea is to utilize the Dynamic Voltage and Frequency Scaling (DVFS) technology in power-scalable clusters, in which the power level will be scaled down when processors are not fully utilized and scaled up when processor workloads are heavy. Intel SpeedStep and AMD PowerNow! are typical examples of DVFS approaches. Researchers in Princeton University investigated the possibility of introducing DVFS technology to interconnections. In addition, researchers from UC San Diego, Duke University, HP labs and Arizona State University proposed a series of cyber-physical approaches to reduce the power cost caused by cooling systems. In recent years, we have proposed several scheduling algorithms for soft or mixed real-time systems, especially for data stream management systems. We also have designed energy-efficient resource management mechanisms for large-scale supercomputing systems, mobile clusters and wireless networks. These algorithms have been published in a number of prestigious journals and conferences. This paper is focused on energy-aware task scheduling on homogeneous cluster servers. We propose a novel heuristic Processor Reduction Optimizing (PRO) algorithm to decrease the number of processors. Simulation results show that the proposed PRO algorithm can significantly reduce energy consumption without performance degradation.