Design and development of hybrid integrated thermal aware job scheduling on computational grid environment
Design and development of hybrid integrated thermal aware job scheduling on computational grid environment
复制标题
计算网格环境下混合集成热感知作业调度的设计与开发
DOI：
10.1109/ismsc.2015.7594020
复制
发表时间：
2015
期刊：
影响因子：
--
通讯作者：
N. Zakaria
中科院分区：
文献类型：
--
作者：
Ahmad Abba Haruna;L. T. Jung;N. Zakaria
几十年来，人们已经看到了从大型主机计算到商品化、现成的高性能计算机集群的变化。目前，数据中心具有数千或数万台高性能计算机，这些计算机为数万或数十万用户提供服务以及计算。传统的IT挑战，如调度，资源分配，现在数据中心正在处理功耗，这是数据中心最昂贵的运营成本因素。低效的冷却导致高温，这反过来又导致硬件故障。本文提出了一种基于基线方法的混合集成热感知调度算法。本文的目的是最大限度地减少冷却能源消耗的数据中心实验室时，分配工作的计算。这些算法避免了高的热应力情况，如大的热点和热违规事件。结果表明，混合集成热感知调度算法（TFCFS和TRR）相比，基线作业调度算法，先来先服务（FCFS）和轮循（RR）减少了7000千瓦的冷却电力。
Over the decades people have seen a change from large mainframe computing to commodity, off-the-shelf clusters of high performance computers. Currently data centers have thousands or tens of thousands of high performance computers that provides services as well as computation for tens or hundreds of thousands of users. Traditional IT challenges such as scheduling, resource allocation and now data centers are dealing with power consumption, it is the most expensive operational cost factor in data centers. Inefficient cooling leads to high temperature and this in turn leads to hardware failure. In this paper a Hybrid Integrated Thermal Aware Scheduling Algorithms are proposed based on baseline approaches. The aim of this paper is to minimize cooling energy consumption in data center labs when assigning jobs for computation. These algorithms avoid high thermal stress situations such as large hotspots and thermal violations events. The results show that the Hybrid Integrated Thermal aware scheduling algorithms (TFCFS and TRR) reduces cooling electricity by 7000KW compared to the baseline job scheduling algorithms, First Come First Serve (FCFS) and Round Robin (RR).