A Quantitative Rule for The use of Resources in a Multiprogrammed ComputerSystem
Edward A. Silver, Alan Loss, Fischer Black · INFOR Information Systems and Operational Research · 1971
The computer has been used extensively to help solve problems in management science, but little has been done in the opposite direction. This paper describes an application ofquantitative techniques to improve the throughput of a computer system.One of the trendsin large-scale computer operations is towards multiprogrammed systems. Here the adjectivemultiprogrammed is meant to imply that two or more jobs simultaneously occupy the main memory of the computer and share resources such as the centra! processor, machine operators, and input/output devices. Under such a parallel configuration, any hardware and/or software change that is geared to speed up the completion of a particularjob may have over-compensating detrimental effects on the other jobs that reside in parallel with the one under consideration.In this paper, a quantitative rule is developed, based on the criterion of maximizing total throughput of jobs, to decide whether or not a proposed change is desirable. Careful consideration is given to how the system parameters and individual program parameters required in the rule would be estimated.