Admission Control for ATM High-Speed Networks

ATM（异步传输模式）高速网络可容纳各种类型的业务，如数字化语音、电影、编码视频和数据等。近年来，许多研究人员报道了ATM高速网络中业务量具有自相似和长期依赖等复杂特性。此外，作为评价这种ATM网络服务质量（QoS）的标准，由于缓冲区溢出导致的丢包概率要求控制在非常小的水平以下，例如10^<-9>的量级。因此，为了保证ATM高速网络的QoS要求，确定罕见事件概率并进行允许控制变得非常重要。在本研究中，我们考虑了由两个并行队列组成的ATM高速网络，这通常用于建模具有两种不同类型流量的通信系统：实时流量（如语音和视频）和非实时流量（如数据）。分析了系统在不同到达过程下的尾部行为。然后，我们利用大偏差技术推导了缓冲区溢出概率的上界和下界，并给出了允许控制算法。我们还利用带保护通道的有限缓冲排队模型对无线移动网络进行了性能分析。具体来说，我们分析了以下几个系统。(1)具有两个类和单个服务器的离散时间轮询系统。到达过程是马尔可夫调制过程，服务调度是伯努利调度。(2)两类多服务器的离散时间轮询系统。到达过程是一般的，服务计划是马尔可夫的。针对这两种轮询系统，我们推导了丢包概率的大偏差上界和下界，并给出了接纳控制算法。（II）对于马尔可夫轮询系统，我们推导了丢包概率的大偏差上界和下界，并将结果应用于无线移动网络。（III）采用有限缓冲排队模型对无线移动网络进行建模，并采用递归算法推导出稳态分布。少

ATM(Asynchronous Transfer Mode) high-speed networks accommodate various types of traffic such as digitized voice, movie, encoded video and data, etc. Recently, it has been reported by many researchers that traffics in ATM high-speed networks have many complicated properties-self-similarity and long-terms dependence, etc. Furthermore, as the criterion for evaluating QoS(quality of service) of such ATM networks, the packet loss probabilities due to the buffer overflow are required to control below very small level, e.g., in the order of 10^<-9>. Thus, determining the rare event probability and performing the admission control become very important in order to provide the guarantee of QoS requirements for ATM high-speed networks.In the present study, we have considered ATM high-speed networks consisting of two-parallel queues, which are generally used in modeling communication systems with two different types of the traffics : real-time traffic(such as voice and video) and non-real-time t … More raffic(such as data). We analyzed the tail behavior of the systems under various arrival processes. Then, we derived the upper and lower bounds of the buffer overflow probabilities using the large deviations techniques, and provided the admission control algorithms. We also carried out the performance analysis for wireless mobile networks by using the finite buffer queueing models with the guard channels. Concretely, we have analyzed the following systems.(I)(1)the discrete-time polling system with two classes and a single server. The arrival processes are Markov-modulated processes and the service schedule is Bernoulli. (2)the discrete-time polling system with two classes and multiple servers. The arrival processes are general and the service schedule is Markovian. For those two polling systems, we derived the large deviations upper and lower bounds of the packet loss probabilities and presented the admission control algorithm.(II)For Markovian polling systems, we derived the large deviations upper and lower bounds of the packet loss probabilities, and applied the results to wireless mobile networks.(III)We modeled wireless mobile networks by using finite buffer queueing models, and derived the steady-state distributions using recursive algorithm approach. Less