Real Time QoS Provisioning in Scalable, Heterogeneous, Unreliable IP Networks

The goal of this effort is the design and implementation of QoS network provisioning strategies that (1) permit real time delivery of QoS services by promptly accepting new calls, and (2) enable prompt recovery (within 100-300 ms) from QoS interruption dueto use/network mobility or network failures. QoS support for internet multimedia applications is necessary to meet stringent end-to-end requirements such as bandwidth, delay and priorities, scheduling, traffic classification. To scale the solution to largenetworks, we used MPLS (Multiprotocol label Switching) and Class aggregation using the Diff Serv model. To support path protection and fast recovery, the notion of spare resource pre-allocation on MPLS paths was introduced. In Phase I many of the precedingbuilding blocks were evaluated via simulation. In Phase II we will implement the proposed QoS framework on a small test-bed to demonstrate its effectiveness in a physical network. A hybrid simulation that supports physical deployments linked withsimulation models of larger networks will be used to demonstrate scalable to a heterogeneous network with a backbone as well as local (wired and wireless) subnets.