Parallel Job Scheduling Using Bacterial Foraging Optimization for Heterogeneous Multi-cluster Environment

Abstract

A variety of clusters are grouped together to form a multi-cluster environment which can tackle the computational needs of a system which cannot be addressed by a single cluster. Studying multi-cluster frameworks is turning challenging day by day as it requires contemporary tools to move alongside with rapidly development and enhanced complexity of one system. Job scheduling in considered as NP hard problem in parallel and distributed computing environments such as cluster, grid and clouds. The way jobs are scheduled by the scheduler is dependent on various factors like number of jobs, processor availability, arrival time etc. Metaheuristics techniques like Genetic Algorithms, Ant Colony Optimization, Artificial Bee Colony, Cuckoo Search, Firefly Algorithm, Bat Algorithm etc. are used by researchers to get near optimal solutions to job scheduling problems. This work addresses a scheduling problem with multiple objectives. The makespan and flowtime are minimized simultaneously solving the issue of optimal job allocation. This work also includes the detailed description of parallel computing and various types scheduling as well as scheduling environments. The performance of the multi-cluster environment is optimized by applying a novel meta-heuristic technique named Bacterial Foraging Optimization Algorithm. This algorithm has better convergence and is not affected by the size of problem. The proposed algorithm was evaluated for different job sets on 3 types of processor configurations. And the final values were compared to those of the existing algorithm. The results show that the proposed algorithm has performed better than the existing one and it can be concluded that the proposed algorithm is feasible and effective for optimal allocation of jobs.