Sensitivity of Rate Control Mechanisms to Uncertainties in Long-range Dependent Traffic Modelling

Sven Ostring

University of Canterbury, New Zealand


There is significant evidence that network traffic is better modelled by self-similar processes, which has implications on the design of network management algorithms. When control mechanisms are designed specifically for the long-range dependent nature of traffic, the issue of the sensitivity of the these controllers to deviations of the traffic from its estimated LRD model becomes apparent. In this seminar, the concepts of self-similarity and long-range dependence will be introduced, together with a rate control mechanism which we have developed. The asymmetric sensitivity with regards to the Hurst parameter, which is predicted by an analytical study of the algorithms, will be shown to occur with actual data traces. Finally, robust control techniques will be discussed which minimise the effect of the asymptote in the sensitivity surface.

[Sven Ostring]
[University of Canterbury, New Zealand]