Task Allocation in Distributed Systems

Abstract

Objectives: In this paper, the response time of the tasks is minimized by migrating the tasks from the overloaded computers to the under load computers using a load balancing technique. Methods/Statistical analysis: Load balancing algorithms belong to the nearest neighbor technique considers only the neighbor computers for migrating the tasks to reduce communication cost between the computers. Though several nearest neighbor techniques are available, this paper uses the diffusion technique for balancing the tasks between the computers in a distributed system, as the strength of the diffusion technique lies in all port communication model and asynchronous implementation. Findings: To evaluate the performance of the system, the load on the system is varied with the mean inter arrival time of the tasks at each computer. The simulation has been carried out to observe the performance of the proposed algorithm (Dynamic Distributed Diffusion) with the existing algorithms SID, GDE, AN by considering tasks arrive to a computer with a Poisson process and follow n* M/D/1 Queuing model. The simulations show the proposed algorithm reduced the load balancing time, compared to the existing algorithms in the literature. Application/Improvements: The proposed algorithm can be suited to any topology and it reduces the load balancing time and hence the response time of the tasks is minimized. Keywords: Diffusion, Load Balancing, Migration, Nearest Neighbor, Response Time, Task Allocation