基于代价优化的云工作流调度改进PSO算法

Cloud Workflow Scheduling Improved Particle Swarm Optimization Algorithm Based on Cost Optimization

摘要:

云计算可以通过即付即用的方式向用户工作流提供资源。为了解决资源服务代价异构环境下的云工作流任务调度代价问题，提出一种基于改进粒子群算法的云工作流任务调度算法WSA-IPSO。通过综合考虑任务的执行代价和依赖任务间发生数据传输时的通信代价，算法将总代价优化问题形式化为有向无环图DAG中的任务调度模型，并提出基于改进粒子群算法的优化模型对其进行求解。通过改进传统粒子群算法的粒子速度更新策略和惯性权重更新策略，算法可以以更快的收敛速度得到代价最小化的调度方案。通过仿真实验，与MCT算法及标准粒子群算法进行性能比较。实验结果表明，WSA-IPSO算法在降低总代价、任务分布的负载均衡以及算法收敛性方面比较同类算法均表现出更好的性能。

Abstract:

To solve the scheduling cost optimization of cloud workflow tasks in dynamic resource service prices environment, a cloud workflow tasks scheduling algorithm based on improved particle swarm optimization WSA_IPSO was proposed. WSA_IPSO considered comprehensively the execution cost of tasks and the communication cost between dependent tasks when they transferred data, formalised the optimization of total cost as a task scheduling model in DAG and presented an improved PSO algorithm to solve. The proposed algorithm was compared with MCT and standard particle swarm optimization algorithm by simulation experiments. The experimental results showed that WSA-IPSO performed better in reducing total cost, load balance of tasks distribution and convergence.