A simulation Study of Dynamic Resource Allocation Scheme for Distributed transactions

Abstract

Most distributed systems nowadays are consists of various nodes having different functions and/or different processing capabilities and speeds. We have considered a heterogeneous distributed system consists of a set of nodes(autonomous computers) with same functionality but different processing capability. This paper is aimed to find out the number of autonomous computer required to design a distributed computing platform (DCP) for a specific problem domain. The transactions initiated in any node routed to the central job scheduler (leader) for execution. The leader once elected is assumed to be fault-free during the assignment and execution of a submitted task. Scheduler model uses Dijkstra algorithm that takes half of that of Bellman-Frod algorithm for Single Origin Shortest Path problem(SOHP). Simulation are being done on a Pentium-II computer system with the model developed using Boroland C++.