Stochastic Greedy-Based Particle Swarm Optimization for Workflow Application in Grid

Abstract

The workflow application is a common grid application. The objective of a workflow application is to complete all the tasks within the shortest time, i.e., minimal makespan. A job scheduler with a high-efficient scheduling algorithm is required to solve workflow scheduling based on grid information. Scheduling problems are NP-complete problems, which have been well solved by metaheuristic algorithms. To attain effective solutions to workflow application, an algorithm named the stochastic greedy PSO (SGPSO) is proposed to solve workflow scheduling; a new velocity update rule based on stochastic greedy is suggested. Restated, a stochastic greedy-driven search guidance is provided to particles. Meanwhile, a stochastic greedy probability (SGP) parameter is designed to help control whether the search behavior of particles is exploitation or exploration to improve search efficiency. The advantages of the proposed scheme are retaining exploration capa- bility during a search, reducing complexity and computation time, and easy to implement. Retaining exploration capability during a search prevents particles from getting trapped on local optimums. Additionally, the diversity of the proposed SGPSO is verified and analyzed. The experimental results demonstrate that the SGPSO proposed can effectively solve workflow class problems encountered in the grid environment.