Geometric Convergence of Finite-Level M/G/1-Type Markov Chains

摘要：

　　This paper studies the geometric asymptotics of the difference between the stationary distribution vectors of a finite-level M/G/1-type Markov chain and its infinite-level extension as the maximum level approaches infinity.This study is intended to evaluate the error of approximate formulae for the loss probability of finite-buffer queues where the queue length processes(possibly embedded at appropriate events)are finite-level M/G/1-type Markov chains.We assume that the level increment is light-tailed,under which we derive a simple asymptotic formula such that the stationary distribution vector of the finite-level Markov chain converges geometrically to that of its infinite-level extension.

关键词： Finite-level M/G/1-type Markov chains Finite-buffer queues Asymptotic analysis Stationary distribution Light-tailed

作者: Tatsuaki Kimura Yosuke Katsumata Hiroyuki Masuyama

作者单位: NTT Network Technology Laboratories NTT Corporation,Tokyo 180-8585 Japan Department of Systems Science,Graduate School of Informatics Kyoto University,Kyoto 606-8501,Japan

会议类型: 国际会议

会议名称: The 12th International Conference on Queueing Theory and Network Applications(第十二届排序理论与网络应用国际会议)(QTNA 2017)

会议地点: 河北秦皇岛

会议语种:英文

页码: 74-77

在线出版日期: 2017-08-21（万方平台首次上网日期，不代表论文的发表时间）