【作者】 金顺福；

【导师】 田乃硕；

【作者基本信息】 燕山大学 ， 电路与系统， 2006， 博士

【摘要】 随着互联网应用的普及,用户对带宽及网络服务质量QoS(Quality of Service)的要求逐步提高。在众多网络控制技术中,IP技术以其灵活、高效的数据交换方式著称,而ATM(Asynchronous Transfer Mode)既可以解决带宽问题,也能够为IP提供很好的QoS,且话音、图像和视频流等多媒体业务也可以应用在ATM上,IP-over-ATM技术应运而生。IP-over-ATM是建立在交换虚通道SVC(Switched Virtual Channel)基础上的,对SVC的性能分析与评价是IP-over-ATM技术研究与发展前提。本文在前人所完成的,基于连续时间排队理论的SVC的性能分析理论的基础上,顺应数字化技术的发展趋势,以离散时间排队理论为基础,依据SVC的运作机制,建立了一套系统的,具有递进关系的,带有启动机制的延迟休假排队模型。通过对排队系统的理论分析,结合数值例子与仿真实验对SVC的性能做了系统的研究。 首先,对前人所完成的,基于连续时间SVC的性能指标分析工作进行了归纳、总结与分析,提炼出了一套较系统的,基于连续时间的SVC性能指标分析的排队模型:带有启动实施/关停延迟的M/G/1排队模型;带有启动实施/关停延迟/关停实施的M/G/1/K排队模型;带有启动实施/关停延迟/关停实施的MAP/G/1/K排队模型;带有启动实施/关停延迟/关停实施的BMAP/G/1/K排队模型。针对每一种模型给出了稳态下SVC建立(释放)比率、SVC有效利用率、SVC的闲置率及信元平均等待时间及丢失率等性能指标解析式。 其次,针对网络中用户触发事务的无后效性,建立了带有启动实施/关停延迟/关停实施的Geom/G/1排队模型,利用嵌入Markov链方法,给出了稳态下到达间隔具有无后效性的信元的平均响应时间、SVC建立(释放)比率、SVC有效利用率、SVC闲置比率等性能指标解析式,通过数值例子和仿真实验,进一步解释了性能指标与超时定时器长度及系统负载大小的关系。 再次,在排队模型的输入过程中引入成批到达机制,建立了带有启动实施/关停延迟/关停实施的Geom~ζ/G/1排队模型,在批量大小服从一般分布的前提下,利用嵌入Markov链方法,给出了排队性能指标。针对信元到达过程的突发性,作为特例,设批量大小ζ服从Pareto(c,α)分布,导出了稳态下到达呈现突发性的信元的平均响应时间、SVC建立(释放)比率、SVC有效利用率、SVC闲置比率等性能指标的解析更多还原

【Abstract】 Along with the popularization of the Internet, user’s requirement for bandwidth and QoS (Quality of Service) is improving. IP technology is famous for its agility and high efficiency in data exchanging among many network controlling methods, while ATM (Asynchronous Transfer Mode) can not only solve the problem of bandwidth, but also provide perfect QoS for IP, in addition, multimedia data, such as audio, image and video flow, can all be transferred on ATM. IP-over-ATM emerges as the times require. IP-over-ATM is based on SVC (Switched Virtual Channel), and the performance analysis and evaluation of SVC is precondition of IP-over-ATM technology research and development. According to SVC’s operation mechanism, paralleling with SVC’s performance analysis on continuous time queue theory, and conforming to digital technologies’ development, a set of systemic discrete time delayed vacation queue models with Setup are built based on discrete time theory. Systemic researches on SVC’s measurement evaluation are made by the method of theoretical analysis together with numerical examples and simulated experiments.Firstly, the work of SVC’s performance analysis accomplished on continuous time are summarized and analyzed, a set of systemic queue models of SVC’s performance analyses based on continuous time are abstracted as follows:M/G/l queue model with Setup/Close-Delay;M/G/1/K queue model with Setup/Close-Delay/Close-Down;MAP/G/1/A: queue model with Setup/Close-Delay/Close-Down;BMAP/G/1/A: queue model with Setup/Close-Delay/Close-Down. SVC performance measurements such as setup(release) ratio, utility ratio, idle ratio and cell’s average waiting time under steady status are given.Secondly, aiming at memoryless property of cell arrival initiated by users, queue model Geom/G/1 with Setup/Close-Delay/Close-Down is built. Under steady state condition, SVC performance measurements such as cell average response time, setup(release) ratio, utilization ratio, and idle ratio, etc. are derived by the method of embedded Markov chain. Dependence relationship between these measurements anddelay timer, system load is shown through numerical examples and simulated experiments.Thirdly, considering cell bursty arrival, batch arrival mechanism is introduced and batch size % is supposed to be Pareto(c,a) distributed, a queue model GeomVG/1 withSetup/Close-Delay/Close-Down is built. Using embedded Markov chain method, analytic expressions of average response time and other SVC measurements such as setup(release) ratio, utility ratio, idle ratio etc. are derived. Furthermore, the influence of bursty arrival on SVC performance is shown by numerical examples and simulated experiments.Finally, considering correlation and bursty property of network traffic, cell arrival process is extended to D-BMAP(Discrete-Time Batch Markov Arrival Process), a queue model D-BMAP/G/1 with Setup/Close-Delay/Close-Down is built. A discrete time factorization principle with general vacation is proposed and proved. In the application with correlated and bursty arrival, steady state performance measurement analytic expressions of average response time, SVC setup(release) ratio, SVC utility ratio, SVC idle ratio etc. are derived, and some numerical examples are made to explain these formulae.更多还原