Particle Swarm Scheduling for Work-Flow Applications in Distributed Computing Environments

Recently, the scheduling problem in distributed data-intensive computing environments has been an active research topic. This Chapter models the scheduling problem for work-flow applications in distributed data-intensive computing environments (FDSP) and

  • PDF / 352,218 Bytes
  • 16 Pages / 439.2 x 666 pts Page_size
  • 41 Downloads / 180 Views

DOWNLOAD

REPORT


2 3

Centre for Quantifiable Quality of Service in Communication Systems, Faculty of Information Technology, Mathematics and Electrical Engineering, Norwegian University of Science and Technology, NO-7491 Trondheim, Norway. [email protected], http://www.softcomputing.net School of Computer Science and Engineering, Dalian Maritime University, 116026 Dalian, China. [email protected] School of Software, Dalian University of Technology, 116620 Dalian, China. mingy [email protected]

Summary. Recently, the scheduling problem in distributed data-intensive computing environments has been an active research topic. This Chapter models the scheduling problem for work-flow applications in distributed data-intensive computing environments (FDSP) and makes an attempt to formulate the problem. Several meta-heuristics inspired from particle swarm optimization algorithm are proposed to formulate efficient schedules. The proposed variable neighborhood particle particle swarm optimization algorithm is compared with a multi-start particle swarm optimization and multi-start genetic algorithm. Experiment results illustrate the algorithm performance and its feasibility and effectiveness for scheduling work-flow applications.

Key words: Particle Swarm Optimization, Scheduling, Work-Flow Applications, Variable Neighborhood Search, Multi-Start Genetic Algorithm, Distributed Computing Environments

13.1 Introduction With the development of the high performance computing (HPC), computational grid, etc., some complex applications are designed by communities of researchers in domains such as chemistry, meteorology, high-energy physics, astronomy, biology and human brain planning (HBP) [1], [2]. For implementing and utilizing successfully those applications, one of the most important task is to find appropriate schedules before the application is executed. The goal is to find an optimal assignment of tasks in the applications with respect A. Abraham et al.: Particle Swarm Scheduling for Work-Flow Applications in Distributed Computing Environments, Studies in Computational Intelligence (SCI) 128, 327–342 (2008) c Springer-Verlag Berlin Heidelberg 2008 www.springerlink.com 

328

Abraham, Liu and Zhao

to the costs of the available resources. However, the scheduling problem in distributed data-intensive computing environments seems quite different from the conventional situation. Scheduling jobs and resources in data-intensive applications need to meet the specific requirements, including process flow, data access/transfer, completion cost, flexibility and availability. All kinds of components in the application can interact with each other directly or indirectly. Scheduling algorithm in traditional computing paradigms barely consider the data transfer problem during mapping computational tasks, and this negligence would be costly in the case of distributed data-intensive applications [3]. Priority scheduling plays a crucial role in the differentiated services architecture for the provisioning of Quality-of-Service (QoS) of network-based applications. Jin an

Data Loading...