An analysis of network performance degradation induced by workload fluctuations

Presents an analysis of network performance degradation induced by workload fluctuations. The basic idea of the analysis is based on catastrophe theory. Conventional performance models typically express workload only by using its average properties. However, it may not be possible to predict network performance simply by using average workload properties, since computer networks are nonlinear systems and workload of computer networks usually fluctuates intensely. The interaction between workload fluctuations and the nonlinearity may greatly affect network behavior. The present analysis shows that fluctuations in workload may change the dynamic structure of the system state space and cause the network performance to vary drastically.<
>