Job scheduling based on job I/O and CPU requirements

In the last few years, there have been efforts to accurately represent a job in terms of a model of its CPU and I/O burst characteristics with the intention that computer operating systems might use this information to improve system throughput. This thesis looks ahead to the time when such burst models may exist and examines some job scheduling concepts and strategies that might be used based on the availability of such burst information. Three reasonable job scheduling algorithms are defined and evaluated via the use of a simulation model of the single CPU, four channel system. The first-come first-served algorithm along with two worse case algorithms are also simulated for comparison purposes. A background in job scheduling concepts and a discussion of job scheduling algorithms currently in use or proposed are also presented.Computer Science, Department o