Distributed Measurement of Packet Loss Probability with Application to Admission Control of VPN Services

When provisioning quantitative QoS assurance for the VPN services over packet switched networks, parameters such as packet loss, delay and delay jitter, besides the required bandwidth, have to be guaranteed. This paper proposes a distributed solution to measure, estimate and control the packet loss probability for the VPN services in the MPLS based core networks. Specifically, the distributed measurement system employs the intelligent agent based architecture to measure the VPN traffic on all of the heterogonous remote nodes. Then the packet loss probability is estimated on-line by applying the large deviation theory. As one of the applications, the hybrid centralized and distributed admission control of new VPN services is also proposed and evaluated on the live network. The numeric results show that the statistical QoS of VPN services, i.e. packet loss probability, can be controlled with the scheme.