A scalable QoS control framework for the next generation IP/MPLS core network

As networks, services and applications converge on a common IP/MPLS backbone, carriers must manage traffic flows in a scalable way with very granular and stringent QoS requirements. In this paper, a control theory approached quantitative QoS control framework is proposed in a distributed and scalable architecture, which assembles the author's previous studies in a complete solution. The dynamic control system is constructed where a reactive estimator is employed for calculation of the packet loss probability, a recursive prediction estimation method is used to identify the parameters of the dynamic packet loss system, and a robust controller is designed for conducting the controller poles in an adaptive manner to keep the packet loss probability in desired limits. The system is designed and implemented with CORBA and J2EE technology in a distributed architecture. The performance is evaluated on the five NCIT*net2 network and the results demonstrate that the system is effective and practical.