摘要
近期对大量实际网络的测量表明,现代网络的业务流特性和部分信源的特性,更适于采用具有长期相关性的自相似或分形模型来描述。已有研究表明长期相关业务的排队特性与基于短期相关模型的有很大不同。但对于自相似业务下ATM网络中复接器性能的研究尚未深入进行。本文在提出一种准自相似(QSSP)的长期相关业务流模型的基础上,求得了N路同参数QSSP输入时复接器的信元丢失率和复接增益,并对非同参数输入的情况提出了一种快速的数值求解方法,用于得到复接器的信元丢失率上限。计算机仿真表明了分析的准确性。
Recent measurements show that the traditional traffic models cannot be success on kinds of long-range dependent traffic which has great influences on the network performance. Fractal or self-similar models are found to be more suitable for describing the modern traffic. But there is still little knowledge about the performance of the multiplexer under self-similar traffic. This paper proposes a Quasi Self-Similar Process (QSSP). Using this process, the upper bond of the cell loss rate and multiplexing gain of the multiplexer in N i.i.d. QSSP input case is gotten. If the sources are with different parameters, an efficient numerical algorithm to get a bond is also proposed. Simulation indicates that the analysis fits with the simulation result well.
基金
国家自然科学基金
电科院预研基金
关键词
长期相关性
复接器
ATM
B-ISDN
Long-range dependence, Self-similar process, Fractal Gaussian noise, Multiplexer, Cell loss rate
