The node-centric formulation for network utility maximization of multihop wireless networks with elastic and inelastic traffic

Abstract

We consider a network with two kinds of traffic: inelastic and elastic. The inelastic traffic requires fixed throughput and high priority, while the elastic traffic has the rate that can be controlled and low priority. Given the fixed rate of inelastic traffic, the problem of how to inject the elastic traffic into the network to achieve the maximum utility of elastic traffic is solved in this paper. The node-centric formulation in cross-layer design framework is applied. By using Lagrange duality method, the congestion control, back-pressure routing and Max-weight scheduling are integrated naturally when decomposing the Larangian. The Greedy Distributed scheduling is introduced to not only decentralize scheduling, but also decrease the complexity of the Max-weight scheduling. The Back-pressure routing has presented the poor performance at the light load, the packets take an unnecessary long route from the source node to the destination node. As the result, the delay is excessively large at light load. The enhancement algorithm is implemented to maximize the utility while minimizing the resource usage. An admission control scheme is also introduced to check if the new inelastic flow is admissible.