Recursive parameter estimation for the dynamic packet loss system

Provisioning QoS enabled MPLS VPN services is increasingly important for network service providers. A series of studies investigate the application of control theoretic frameworks to measure and control the QoS parameters, especially the packet loss probability. The plant model is one of the key issues to be solved in the control system. This paper will try to identify the dynamic model of packet loss probability using the system identification methods, where the traffic rates are inputs and packet loss probability is the output. The experiments are designed on the live NCIT*net2. The numeric results are evaluated to verify the effectiveness of the algorithm.