Balancing processor shares of scheduling classes through controlled allocation of memory

  • Authors:
  • K. V. Sastry

  • Affiliations:
  • Sperry Univac, Roseville, Minnesota

  • Venue:
  • AFIPS '80 Proceedings of the May 19-22, 1980, national computer conference
  • Year:
  • 1980

Quantified Score

Hi-index 0.00

Visualization

Abstract

In a paged, virtual memory computer system used by several different classes of users simultaneously, it is reasonable to expect that each class will demand certain service rates. Service rates for a particular class can be affected by having each class use its share of (a specified percentage) Instruction Processor (IP) time at regular intervals. There are also two other important resources in the system---main memory and the IO system. The service rate of a class is very much dependent upon the availability of (or lack of) these two resources even if a processor is dedicated to a class. On the other hand, a specified IP share can be achieved by using an "unfairly large" portion of main storage for an "unfairly large" population of processes in the READY state. Although it may be favorable to do so for a particular class, it may, in fact, prevent the other classes from achieving their processor shares because of lack of enough storage.