Measurement and Control of Packet Loss Probability for MPLS VPN Services

摘要

Multiprotocol-label-switch virtual private network (VPN) service has emerged as a value-added cost-effective VPN- based service, and its market opportunity is tremendous for network service providers (NSPs). Quality of service (QoS) control is one of the critical issues to be addressed, for NSPs to maximize their profits. This paper proposes a QoS control system that combines the service admission control (SAC) and the rate feedback control together to maintain the preset QoS parameters (the packet loss probability case) in the provider's backbone network. First, one measurement module employs distributed intelligent agents to measure the VPN traffic from all of the heterogenous remote network nodes. The packet loss probability is then estimated on line by applying the large-deviation theory. SAC strategy decides the amount of permitted VPN services, while the feedback controller dynamically throttles ingress traffic rates of instantiated VPN services to maintain the preset packet loss targets. The performance is evaluated in the live network of the National Capital Institute of Telecommunications (NCIT$ast$net2). Numeric results show that, under appropriately selected control parameters, it is possible to maintain the network operation within a prescribed loss limit.