Multi-objective Evolution-Based Scheduling of Computational Intensive Applications in Grid Environment
Grid computing has been evolved as a high performance computing to fulfill the demand of computational resources among the geographically dispersed virtual organizations. Grid is used to provide solutions to the complex computational intensive problems. S
- PDF / 379,997 Bytes
- 11 Pages / 439.37 x 666.142 pts Page_size
- 45 Downloads / 183 Views
Abstract Grid computing has been evolved as a high performance computing to fulfill the demand of computational resources among the geographically dispersed virtual organizations. Grid is used to provide solutions to the complex computational intensive problems. Scheduling of user applications on the distributed resources is an indispensable issue in Grid environment. In this paper, a speedconstrained multi-objective particle swarm optimization (SMPSO) technique-based scheduler is proposed to find efficient schedules that minimizes makespan, flowtime, resource usage cost and maximizes resource utilization in Grid environment. The work is integrated in ALEA 3.0 Grid Scheduling simulator. The results of the proposed approach have been contrasted with Grid’s conventional scheduling algorithms like FCFS, EDF, MinMin, and other multi-objective algorithms like NSGA-II and SPEA2. The results discussed in the paper shows that SMPSO outperforms over other scheduling techniques. Keywords Grid scheduling Multi-objective SMPSO
⋅
⋅
Resource utilization
⋅
Makespan
⋅
Flowtime
⋅
1 Introduction Grid is a kind of distributed computing where resources are owned by multiple virtual organizations and follows common rules for sharing and selection of resources to execute the user jobs from diverse locations [1, 2]. Grid middleware allows the users to interact with Grid systems. The concept of computational grid has been evolved from power Grid. The computational grids are required to provide reliable and low-cost methods for sharing resources such as computers, storage and working memory, network bandwidth, and software applications across geographically distributed virtual organizations [3]. The resource providers advertise Mandeep Kaur (✉) Computer Science Department, Savitribai Phule University, Pune, India e-mail: [email protected] © Springer Science+Business Media Singapore 2017 S.C. Satapathy et al. (eds.), Proceedings of the International Conference on Data Engineering and Communication Technology, Advances in Intelligent Systems and Computing 469, DOI 10.1007/978-981-10-1678-3_44
457
458
Mandeep Kaur
their resources on Grid using Grid information service and resource consumers execute their jobs on the advertised and available resources with the help of Grid resource broker. In Grid scheduling, user jobs are to be scheduled on resources that belong to distinct administrative domains. Grid scheduling involves in gathering of information about resources that are capable enough to execute the user jobs, advance reservation of resources, submission of user jobs, and monitoring of user jobs [4]. The Grid scheduler is responsible for sending a user job to the best suitable Grid resource [5]. The grid scheduler matches the optimum resource to the user application. The objective of grid scheduling is to deliver QoS requirements of the grid users like minimum resource usage cost, job execution before predefined deadline and to raise the resource utilization [6].
1.1
Purpose of Research
The motivation behind this rese
Data Loading...
