PSO+LOA: hybrid constrained optimization for scheduling scientific workflows in the cloud

Abstract

Cloud computing provides a cost-effective deploying environment for hosting and executing workflows as its elasticity, scalability and pay-per-use model. Scientific applications are normally compute- or resource-intensive, and how to run them in the cloud while both meeting QoS of users and guaranteeing the benefits of cloud service providers (CSPs) is still a challenge and depends mainly on workflow scheduling. In this article, we propose a hybrid optimization approach, PSO+LOA, i.e., a combination of particle swarm optimization (PSO) and lion optimization algorithm (LOA) for scheduling workflows in the cloud to minimize the total execution time under budget constraints. The main contributions of our work are: (1) A Euclidean distance (ED) aware particle reposition strategy is defined for two close particles, so as to separate them away from each other, hence enhancing the capability of escaping from local optima. (2) To improve the search and convergence efficiency of original PSO, we modify the velocity update equation by introducing adaptive parameters. (3) Inspired by the multiple-swarm co-evolutionary mechanism of LOA, we integrate PSO with LOA to make a good balance between exploration and exploitation during the whole optimization process. Extensive experiments are conducted over well-known scientific workflows with different sizes and types through WorkflowSim. The experimental results demonstrate that in most cases, PSO+LOA outperforms the existing algorithms in the extent of budget constraint satisfiability, solution quality, i.e., it can generate much better solutions which meet the needs of different budget constraints, especially for large-scale applications, such as the average relative deviation index for PSO+LOA and genetic algorithm (GA) are 0.03% and 0.20%, respectively.