An approach to estimating the traffic of a dynamic packet-switched network

To control the performance of a packet-switched network efficiently, some knowledge on dynamic behavior of the network needs to be investigated. In this paper, we apply an adaptive control technique, recursive least square (RLS) estimation, to propose a new approach to estimating the traffic parameters of the network. Based on an introduced dynamic model of the network, this approach can, through adaptively adjusting forgetting factors, not only filter out the noise from estimates while the network operates in a steady state, but also track parameter variations while in a transient state particularly following a step change on the parameter. Since the mean values of the parameters are mainly considered, the dynamic model is suitable to any type of traffic. Therefore, the methodology presented in this paper can then be applied to the issues of B-ISDN network control, such as routing problem and congestion control