Affiliations: [a] Department of Mathematics and Computer Science, Faculty of Science, Menoufia University, Menoufia, Egypt | [b] Department of Computer Science, Faculty of Science, Minia University, Minia, Egypt | [c] College of Computers and Information Technology, Taif University, Al Hawiyah, Kingdom of Saudi Arabia. E-mails: [email protected], [email protected], [email protected]
Correspondence:
[*]
Corresponding author: S.F. El-Zoghdy, Department of Mathematics and Computer Science, Faculty of Science, Menoufia University, Menoufia, Egypt. E-mails: [email protected], [email protected].
Abstract: This paper proposes a dynamic workload’s threshold-based scheduling and load balancing algorithm for distributed computing systems. It takes into account the dynamicity and heterogeneity of available system resources, and tasks. It automatically updates the workload thresholds upon any significant change of any resource workload and balances the system’s workload using the new thresholds. Unlike most of the current scheduling and load balancing algorithms, the proposed algorithm uses the resource’s load as a workload index instead of number of tasks. The number of tasks is not a good workload index especially when the weights of tasks are different. A node can be overloaded even if it has a number of tasks less than that in an under-loaded node. A simulation model is built to evaluate the performance of the proposed algorithm. The performance of the proposed algorithm is evaluated and compared with that of Min–Min, and Ant Colony scheduling algorithms using mean task response time, average utilization of processing nodes, and load balancing level as performance metrics. The results show that the proposed algorithm reduces the task mean response time, improves resources utilization, and load balancing level compared with the other two studied algorithms in the range of system’s parameters values examined.
Keywords: High performance distributed computing systems, scheduling and load balancing, performance evaluation
