Calculation of steady-state probabilities of MM queues further approaches英文文献资料.docVIP

JOURNALOFAPPLIEDMATHEMATICSANDDECISIONSCIENCES,6(1),43–50 Copyright c 2002,LawrenceErlbaumAssociates, Inc. Calculation of Steady-State Probabilities of M/M Queues: Further Approaches DAVID K. SMITH? School of Mathematical Sciences, University of Exeter, Exeter EX4 4QE United Kingdom Abstract. This paper draws attention to the risk of rounding error in the numerical evaluation of steady-state probabilities for the M/M family of queues. A method for avoidingtheriskispresentedwhichiseasytoprogramforcalculationinpractice. Keywords: Accuracy,computing,queuemodels. 1. Introduct ion The M/ M family of queues is generally the ?rst to be introduced to student s of operat ional research and queueing theory. The assumpt ions made in the model are that customers arrive randomly at a service point at a known rat eλ per unit time, wait unt il a server is available and then require a ran- dom time with an exponent ial distribut ion with mean 1/μ for service. It is possible to calculat e the steady state probabilit ies for the number of cus- tomers in the queue system for many variant s of this family of queues, and the result s appear in many textbooks. However, little notice appears to have been taken of the pract icalities of calculat ing these probabilit ies numerically. Recent ly, Pasternack and Drezner([2]) present ed a recursive approach which avoided some compu- tational di?cult ies. In this note, we suggest a di?erent approach which gives similar comput ational accuracy, and is easy to use by student s who are familiar with spreadsheet s. Our interest in the problem arose when circumst ances were encount ered where the obvious approaches to solving the queueing models proved liable to severe rounding error. ? Requests for reprints should be sent to David K. Smith, School of Mathematical Sciences,UniversityofExeter,ExeterEX44QEUnitedKingdom  44 D. K. SMI