Data center task scheduling through Biogeography-Based Optimization model with the aim of reducing makespan

Abstract

Due to the rapid growth in the number of cloud users and the increment of data center users as the basis of clouds thereof, an optimal task scheduling problem would emerge as a vital issue in near future. Since, the complexity of optimal task scheduling nature, which is NP-Complete, the evolutionary algorithms render better performance than simple gradient-based algorithms. In the proposed approach, an evolutionary algorithm based on Biogeography-Based Optimization is applied to achieve optimal task scheduling in data centers. Workloads are distributed over virtual machines in a manner that total execution time (makespan) is minimized. An Information Base Repository (IBR) is considered and applied in order to store the online Virtual Machines load status. The IBR and the workloads information submitted to the data center are applied first to draw decisions for choosing which one of the VMs will be the receptive of the submitted workload; next, forwards the workload to the specified VM. The VM available resources of Memory, Bandwidth, storage and VM CPU Million Instruction Per Second are considered to find the optimal dispatching solution. Simulation results indicate that an increase in the number of VMs, would not change the time of getting optimal solution in a drastic manner and the covergence time increases in a slow graduation compared with task scheduling approaches, which is based on Genetic Optimization and Particle Swarm Optimization. So the total workload will be distributed in an optimal manner.