Task Allocation in Distributed Systems using a Stochastic Programming Approach

Abstract

The rapid progress in microprocessors and the decreasing cost of hardware have

made the distributed proccssmg systems economically allractive for many applications. llowever, many problems in the distributed systems arc still in their development stage.

A distributed processing system consists of a set of processors (processing elements) which are connected via communication channels whose tasks (distributed task) are coordinated by a distributed operating system that controls information processing and flow among processors to achieve the desired requiren,ents.

Task to be processed by the distributed system is decomposed into a set of cooperating tasks ( sub-tasks ) which are then allocated to the system's processors.

Assuming that the distributed task have been partitioned into sub-tasks, finding the optimal allocation for these sub-tasks on the system's processors using goal programming methodology is the aim of this thesis. Two allocation models are included. The first is the deterministic sub-task allocation model in which all model's parameters are assumed deterministic with known values. The second is more general which is the stochastic sub-task allocation model in which some of the model's parameters are assumed random with known distributions.