A scalable approach to the partition of QoS requirements in unicast and multicast

Supporting Quality of Service (QoS) in large-scale broadband networks poses major challenges, due to the intrinsic complexity of the corresponding resource allocation problems. An important problem in this context is how to partition QoS requirements along a selected topology (path for unicast, tree for multicast). As networks grow in size, the scalability of the solution becomes increasingly important. This requires to devise efficient algorithms, whose computational complexity is less dependent on the network size. In addition, recently proposed precomputation-based methods can be employed to facilitate scalability by significantly reducing the time needed for handling incoming requests. We present a novel solution technique to the QoS partition problem(s), based on a "divide and conquer" scheme. As opposed to previous solutions, our technique considerably reduces the computational complexity In terms of dependence on network size; moreover, it enables the development of precomputation schemes. Hence, our technique provides a scalable approach to the QoS partition problem, for both unicast and multicast. In addition, our algorithms readily generalize to support QoS routing in typical setings of large-scale networks.