Network Utility Maximization: A Rate-Distortion Perspective

Network utility maximization (NUM) represents a vast and growing body of literature in optimizing network operation such as throughput and fairness, given a set of constraints. This framework has resulted in a better understanding of optimal operation of and interaction among layers of the protocol stack, including congestion control, routing, access and physical layer transmission. However, traditional NUM optimization does not incorporate lossy compression (rate-distortion) into its formulation - data is assumed pre-compressed and packetized prior to analysis. Since rate-distortion has a substantial impact on end-user experience (for example, in video/multimedia delivery), this paper generalizes the traditional NUM framework to include compression control. It develops a distributed compression control for binary sources, and solves the coupled NUM problem in special cases to illustrate important aspects of compression control. Finally, this paper discusses a stochastic framework that includes compression control, and provide insights on adaptive control of networks.