实时计算机网络的有效带宽

Effective Bandwidths in Real-time Computer Networks

文章通过剖析实时计算机网络研究中“有效带宽”定义的内涵与外延,阐明了它在介质访问仲裁、逻辑链路控制、确定性排队、保证概率排队等应用领域中的意义与用法;并指出了根据大偏离原理(LargeDeviationPrinciple,LDP)定义的有效带宽不仅从概念上反映了多时间尺度和空间尺度下网络的统计复用情况,而且具有带宽的量纲。通过有效带宽与流量过程的累积量生成函数(CumulantGeneratingFunction,CGF)的关系,给出了将大偏离原理应用于流量整形变长数据包流量的方法,并根据随机序(stochasticordering)理论提出了一种“4状态7参数”模型作为这种整形数据流的随机性能界限。通过对CGF函数在空间尺度上求极限,得到最大流率函数,揭示了LDP框架下,保证概率(guaranteeingprobability)意义下到最坏条件下(theworstcase)实时性能保证具有一致的语义。

This paper illustrates the intension and extension of the terminology'effective bandwidth' in the context of real-time network accessing control,logic link control and deterministic queuing etc.,and then demonstrates the definition derived from the Large Deviation Principle(LDP) is not only a function that characterizes the traffic features and statistic multiplexing under real-time communication constraints with respect to multiple temporal and spatial scales,but also has the same unit as bandwidth(link rate).It emploies the LDP to calculus guaranteeing probability of the shaped variable length packet multiplexing by transforming its effective bandwidth from arrival processes' cumulant generating functions(CGF).A'4-state-7-parameter'model is given as a stochastic bound to approximate the crude CGF in the meaning of stochastic ordering.The limits of groups of CGFs with respect to the spatial scales deduce the Maximum Rate Function,which is used in hard real-time analysis.The limit relationship indicates the coherent meaning from probabilistic guarantees to the worst case real-time analysis within the effective bandwidth and the LDP framework.