集群系统任务调度算法  

On task scheduling algorithms in cluster system

作　　者：王文豪[1] 周静波[1] 高尚兵[1] 蒋道霞[2] 

机构地区：[1]淮阴工学院,江苏淮安223003 [2]江苏财经职业技术学院,江苏淮安223003

出　　处：《南昌大学学报（理科版）》2016年第4期403-408,共6页Journal of Nanchang University(Natural Science)

基　　金：国家自然科学基金资助项目(61402192);江苏高校自然科学研究计划资助项目(14KJB520006);江苏省淮安市科技支撑计划资助项目(HAG2013068)

摘　　要：任务调度是集群系统的关键技术之一,针对总线结构DAG模型下的任务调度问题,提出一种基于混沌改进的遗传模拟退火算法,简称Chaos-GSA。该算法在原有遗传模拟退火算法的基础上引入混沌系统,改进种群初始化方法和交叉、变异算子,提高算法的收敛速度,并在降温时考虑染色体资源平均利用率,使具有较大资源利用率的个体更容易被选择。实验结果表明:该算法与传统GSA算法相比,有明显的优越性,可以减少时间跨度,提高资源的利用率。Task scheduling is one of the key technologies in the cluster system.For the task scheduling problem in DAG model of bus structure,an improved genetic simulated annealing algorithm based on Chaos （Chaos-SA）was put forward.As an extension of the existing genetic simulated annealing algorithm,the Chaos-GSA algorithm embedded chaotic systems to create better initial data, refined crossover and mutation operators,yielding improved convergence rate.In the process of the annealing,the factor of the average rate of resources was fully considered,so that the larger rate of resources of individuals,the more likely they were selected.The simulation results showed that the algorithm,compared with the traditional GSA,had obvious advantages of reducing the makespan and improving the usage of resources.

关 键 词：集群系统 任务调度 GA算法 SA算法 混沌遗传退火算法 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]