DECENTRALIZED CONTROL OF DISTRIBUTED PROCESSING SYSTEMS
Ahmed K. Ezzat · University of New Hampshire Scholars Repository (University of New Hampshire at Manchester) · 1982
This thesis presents a methodology for implementing decentralized scheduling for distributed systems. The environment in which the controlling entities make decisions is stochastic and can be described as uncertain since each entity may have a different view of the system state. As a consequence, these entities may make inconsistent decisions. The methodology is based on defining the system state as a set of distributions and using a queueing model to predict the future behaviour of the system. The predicted state is used to schedule the individual job tasks based on minimum predicted job response time. A hypothetical real system is simulated. The methodology was tested using different queueing models and under different environments. An evaluation of the proposed technique using the simulation results indicates a consistent performance improvement over the no network case. Suggestions for extending this research are also presented.