Feedback Queueing Models for Time-Shared Systems

ABSTRACT

Time-shared processing systems (e.g. communication or computer systems) are studied by considering priority disciplines operating in a stochastic queueing environment. Results are obtained for the average time spent in the system, conditioned on the length of required service (e.g. message lenght or number of computations). No chage is made for swap time, and the results hold only for Markov assumptions for the arrival and service processes. Two distinct feedback models with a single quantum-controlled service are considered. The first is a round-robin (RR) system in which the service facility processes each customer for a maximum of q sec. If the customer's service is completed during this quantum, he leaves the system; otherwise he returns to the end of the queue to await another quantum of service. The second is a feedback (FBN) system with N queues in which a new arrival joins the tail of the first queue. The server gives service to a customer from the nth queue only if all lower numbered queues are empty. When taken from the nth queue, a customer is given q sec of service. If this completes his processing requirement he leaves the system; otherwise he joins the tail of the (n + 1)-st queue (n = 1, 2, · · ·, N - 1). The limiting case of N → ∞ is also treated. Both models are therefore quantum-controlled, and involve feedback to the tail of some queue, thus providing rapid service for customers with short service-time requirements. The interesting limiting case in which q → 0 (a “processor-shared” model) is also examined. Comparison is made with the first-come-first-served system and also the shortest-job-first discipline. Finally the FB∞ system is generalized to include (priority) inputs at each of the queues in the system.

REFERENCES

• 1: COBHAM, A. Priority assignment in wMting-line problems. Oper. Res. 2 (Feb. 1954), 70-76.
• 2: COFFMAN, E. G. Stochastic models of multiple and time-shared computer operation. Rep. No. 66-38, Dep. of Engineering, U. of Calif. at Los Angeles, June 1966.
• 3: FELLER, W. An Introduction to Probability Theory and Its Applications. Wiley, New York, 1957.
• 4: KLEINROCK, L. Analysis of a time-shared processor. Nay. Res. Logislics Quart. 11, 10 (March 1964), 59-73.
• 5: --. A conservation law for a wide class of queueing disciplines. Nay. Res. Logistics Quart. 1, 2 (June 1965), 181-192.
• 6: Leonard Kleinrock, Time-shared Systems: a theoretical treatment, Journal of the ACM (JACM), v.14 n.2, p.242-261, April 1967 [doi>10.1145/321386.321388]
• 7: B. Krishnamoorthi , Roger C. Wood, Time-Shared Computer Operations With Both Interarrival and Service Times Exponential, Journal of the ACM (JACM), v.13 n.3, p.317-338, July 1966 [doi>10.1145/321341.321342]
• 8: MILLER, L. W., AND SCHRAGE, L .E . The queue MG1 with the shortest remaining processing time discipline. Rep. P-3263, RAND Corp., Santa Monica, Calif., Nov. 1965.
• 9: PHILIPS, T.E. Machine repair as a priority waiting-line problem. Oper. Res. 9 (Sept.- Oct. 1961), 732-742.
• 10: SCHERR, A. L. An analysis of time-shared computer systems. Ph.D. Diss., MIT, Cambridge, Mass., June 1965.
• 11: SCHRAG, L. E. The queue M/G/1 with feedback to lower priority queues. Manage. Sci. (to be published).
• 12: TAKACS, L. Introduction to the Theory of Queues. Oxford U. Press, New York, 1962.

INDEX TERMS

Primary Classification:
  G. Mathematics of Computing
  G.3 PROBABILITY AND STATISTICS
      Subjects: Queueing theory

Additional Classification:
  D. Software
  D.4 OPERATING SYSTEMS
      D.4.1 Process Management
          Subjects: Multiprocessing/multiprogramming/multitasking

General Terms:
Measurement, Performance, Theory, Verification