蚁群算法在服务器集群批量任务调度中的应用

Application of Ant Algorithm for Scheduling Multi-task in Service Cluster

服务器集群中的负载均衡和作业调度是影响系统性能的重要因素.本文描述服务器集群批量任务的作业调度问题,对该问题建立了基于图的模型.由于使用一般的启发式算法或动态规划算法解决该问题具有局限性,本文引入蚁群算法进行求解,并针对该问题具体求解提出了启发式距离合适的计算方法.最后在仿真的基础上,讨论了算法的优化效果和收敛性,结果表明蚁群算法解决该问题具有优异的性能.

When construct service clusters,the fact of load balancing and task scheduling is very important,it influences the performance of the whole network system.Sometimes,the task scheduling in a system is presented to be NP problem.This paper represents the problem of scheduling multi-tasks in service cluster and builds a graph based model to describe it.Due to the difficulties of resolving this problem using heuristic algorithm and dynamic programming algorithm,the ant colony algorithm,an evolutionary algorithm,is proposed.Then a novel approach of calculating the heuristic distance is proposed when using the ant colony algorithm.After all,based on the simulation,this paper compares the performance between the ant colony algorithm and the traditional round-robin algorithm in resolving this problem,the convergence of the ant colony algorithm is also discussed.The ant colony algorithm is demonstrated to be high performance in resolving the multi-task scheduling problem in service cluster.