Highly advanced cloudlet scheduling algorithm based on Particle Swarm Optimization

Abstract

Cloud computing has become buzzword today. Here, dynamically scalable services and virtualized resources are provided over the internet. Cloud provides following types of services: Software-as-a-Service(SaaS), Platform-as-a-Service(PaaS) and Infrastructure-as-a-service(IaaS). This paper has prime focus on IaaS clouds which offer virtual pool of infinite, heterogeneous resources to users, that can be access on demand. The very objective of this paper is to dynamically optimize task scheduling at system level as well as user level. This paper relates advanced heuristic optimization technique i.e. Particle Swarm Optimization (PSO) which performs better computational over other evolutionary algorithm and optimization technique. In proposed algorithm, group of tasks are represented as particles, where each particle tries to reach its best position in the swarm (collection of all groups of task available for scheduling). Direction and magnitude of velocity can be performed in either of two ways: deadline based or cost based velocity updation, which altogether contributes the particle to reach its local maxima, finally reach its global maxima. This means all the available tasks are efficiently scheduled to the very best of its optimization. We recompile the cloudsim and simulate the proposed algorithm and results of this algorithm are compared with sequential task scheduling. The experimental results indicates that proposed algorithm has higher accuracy in terms of least execution time and at task execution at lower cost that considers heterogeneous resources and elasticity of IaaS clouds that can be dynamically acquired on pay-per-use basis. This algorithm is not only beneficial to user and service provider, but also provides better efficiency i.e. benefit at system level.