QUEUE LENGTH DISTRIBUTIONS FROM PROBABILITY GENERATING-FUNCTIONS VIA DISCRETE FOURIER-TRANSFORMS

被引:17
作者
DAIGLE, JN
机构
[1] Virginia Polytechnic Inst &, State Univ, United States
关键词
Computer Programming--Algorithms - Mathematical Transformations--Fourier Transforms;
D O I
10.1016/0167-6377(89)90066-7
中图分类号
C93 [管理学]; O22 [运筹学];
学科分类号
070105 ; 12 ; 1201 ; 1202 ; 120202 ;
摘要
We review the relationship between the Discrete Fourier Transform (DFT) coefficients and the probability masses. We demonstrate that choosing a larger value for K does not necessarily improve the quality of the results when the objective is to obtain mass probabilities in queueing systems; instead, choosing an unnecessarily large K simply leads both to longer computation times and increased round-off error. We propose an alternative approach which takes advantage of the fact that the tail probabilities in queueing systems decay geometrically. Our approach addresses aliasing error and round-off error simultaneously and, in addition, provides tail probabilities.
引用
收藏
页码:229 / 236
页数:8
相关论文
共 22 条