Dynamic quality of service support in virtual private networks

Abstract

This thesis presents a framework for the provision of dynamic Quality of Service (QoS) support in Virtual Private Networks (VPNs). Specifically, we focus on MPLS based VPN, with the QoS components including use of MPLS Diffserv, MPLS Traffic Engineering, RSVP-TE signaling protocol etc. The framework consists of a Dynamic Bandwidth Allocation model that includes traffic estimators and the development of a resource reservation technique. The Dynamic Bandwidth Allocation (DBA) uses traffic estimators whose outputs are used for the reservation of resource within some time duration in the future. Three traffic estimation algorithms are implemented and tested. The Dynamic Classed Based Queuing (CBQ) technique is applied in our systems to allocate resources. This model can automatically adjust to the bandwidth size of a VPN tunnel based on how much traffic is flowing through the tunnel. An Internet Service Provider (ISP) can simplify the task of managing its network and reduce costs by using our DBA mechanism, taking advantage of the available tunnel bandwidth while still providing guarantees for high-priority traffic. We implemented and evaluated this model on our MPLS Diffserv enabled Linux test-bed. Performance evaluation shows a higher resource utilization can be achieved by using our model.