Distributed network resource allocation for multi-tiered multimedia applications

The continuously growing number of multimedia applications in current communication networks highlights the necessity for an efficient resource allocation mechanism to capture the unique characteristics of multi-tiered multimedia applications and allocate network capacity in an efficient way. This paper examines the problem of sharing the network throughput under the existence of inelastic traffic flows that follow a multi-tiered utility function. First, the concept of multi-sigmoidal utilities is introduced in order to describe user satisfaction, then, the implications of the use of such utilities are discussed for two different allocation policies; the bandwidth-proportional and the utility-proportional fairness allocation policies. In the former case, the intrinsic reasons of possible network oscillations are analyzed in detail and a heuristic to overcome such situations is proposed. In the latter one, where such oscillations are not possible, efficient ways to calculate a closed form solution for the optimal rate allocation are described. Moreover, a novel mathematical representation of such a multi-sigmoidal utility is presented and closed form solutions for a number of application types are calculated. Finally, the efficiency and robustness of the proposed algorithms is evaluated by simulations for different network topologies and compared against other work in literature.