Authors: Lovejit Singh; Sarbjeet Singh
Addresses: Computer Science and Engineering, University Institute of Engineering and Technology, Panjab University, Chandigarh 160014, India ' Computer Science and Engineering, University Institute of Engineering and Technology, Panjab University, Chandigarh 160014, India
Abstract: Cloud service providers offer services to customers through the internet. They use economic models to charge customers for using their resources. Many applications require workflow processing for their execution. In workflow processing, there are dependencies among various tasks, and parent tasks are executed before child tasks. There are various criteria on the basis of which resources are allocated to workflow applications such as time, cost, reliability etc. In this paper, a score-based genetic algorithm is proposed which allocates resources to workflow application tasks by taking into consideration three important factors: score of the machine, makespan and cost. The algorithm allocates those virtual instances to workflow application tasks that result in meeting user-defined deadlines as well as budget. It also reduces the failure rate of workflow applications using a score manager. The performance of the score-based genetic algorithm has been compared with a simple genetic algorithm on the basis of time, cost and failure rate. Different workflows such as CyberShake, Montage, SIPHT and Epigenomics have been executed to evaluate the performance of both scheduling algorithms. The results indicate that the score-based genetic algorithm performs better than the simple genetic algorithm in terms of reducing makespan, failure rate and cost.
Keywords: cloud computing; workflow applications; genetic algorithms; workflow scheduling; makespan; failure rate; cost reduction.
International Journal of Grid and Utility Computing, 2016 Vol.7 No.4, pp.272 - 284
Received: 08 Feb 2015
Accepted: 26 Nov 2015
Published online: 14 Dec 2016 *