一种面向网络边缘任务调度问题的多方向粒子群优化算法

A MULTI-DIRECTIONAL PARTICLE SWARM OPTIMIZATION ALGORITHM FOR NETWORK EDGE TASK SCHEDULING

任务调度是云计算及网格计算环境中的重要问题,已有的调度算法往往仅致力于最小化任务的总执行时间而不设置其他约束条件,以致难以实现多种性能指标的同时优化。所提出的面向网络边缘任务调度问题的多方向粒子群优化算法,用于解决并发任务在网络边缘服务节点中的分布式调度问题,调度的目标是在任务执行的资源开销不超过阈值的情况下,最小化任务完成的总时间。该方法与现有的离散粒子群优化算法相比同时降低了任务的总完成时间及资源开销,且在合理预设资源开销上限的情况下,其计算复杂度实现了较大程度优化。仿真表明,所提出的方法比现有的离散粒子群优化算法的任务总完成时间缩短约10.52%~13.23%,资源开销减少约10.32%~13.29%。同时,在合理降低资源开销阈值的情况下,该方法的程序运行时间比现有的粒子群调度方法明显缩短。

Task scheduling is an important problem in cloud computing and grid computing environments. Existing scheduling algorithms are often only dedicated to minimizing the total execution time of tasks without setting other constraints, making it difficult to optimize multiple performance metrics simultaneously. The proposed multi-directional particle swarm optimization algorithm for network edge task scheduling problem solves the distributed scheduling problem of concurrent tasks in network edge service nodes. The goal of scheduling is to minimize the total time of task completion in the case that the resource cost of task execution less than the threshold. Compared with the existing discrete particle swarm optimization algorithm, this method reduces the total task completion time and resource cost, and achieves a large degree of optimization of its computational complexity in the case of a reasonable preset resource overhead. The simulation results show that compared with the existing discrete particle swarm optimization algorithm, this method can reduce the total task completion time by about 10.52%~13.23% and the resource cost by 10.32% ~13.29%. Meanwhile, the running time of this method is significantly shorter than that of the existing discrete particle swarm optimization algorithm in the case of a reasonable reduction of resource overhead threshold.