Analysis of Two Time-Sharing Algorithms Designed for Limited Swapping

Abstract

The necessity for swapping in the operation of modern time-sharing systems constitutes the major reason for the latter's inefficiency compared to batch-processing systems. Time-sharing algorithms are discussed which are designed primarily for the reduction of swapping without intolerable changes in the waiting time distributions. A particular class of such algorithms in which conventional procedures are modified by making the quantum allocation dependent on input activity is given a more detailed treatment. In particular, queueing models corresponding to these algorithms are devised and then analyzed for the purpose of obtaining the mean waiting times conditioned on the service required. These results are then compared to those obtained for the conventional models and the comparison subsequently measured against the swapping requirements of the two classes of algorithms.