A comparative investigation into the effects of uncertainty in job times on the performance of the shortest processing time and the Carroll C over T algorithms for dispatching computer jobs in a multiprogrammed batch computer

Includes bibliographical references (pages 64-65)Computers have unique characteristics in mode of operation that distinguish them from machines in the classical production management sense. The use of two scheduling techniques from production management for dispatching computer jobs on a batch computer is investigated. The information needed for scheduling purposes on a computer is not always known with certainty prior to jobs being processed. The thesis explores the application of theoretical scheduling techniques under different conditions in which the scheduling data is uncertain. The thesis investigates the impact of different levels of uncertainty with the data. Impact is measured on average response time. The scheduling techniques selected for optimizing this measurement criterion are Shortest Processing Time (SPD and Cost Over Time (COVERT). The performances of these schedulers are compared by examining their average response times obtained for the different levels of uncertainty with the data. These times are compared with the performance obtained by a completely random scheduler such as first in/first out (FIFO). They are also compared with schedules that could have been constructed had the available information been-perfect. The methodology of problem solution includes discrete simulation of a number of models. A description of these models is presented. The results demonstrate that use of theoretical scheduling techniques would be beneficial for computer operations despite conditions of considerable uncertainty which typify the normal computer operation and scheduling