Distributed Margin Optimization Using Spectrum Balancing in Multi-user DSL Systems

Optimizing the SNR margin in Dynamic Spectrum Management (DSM) systems can significantly improve stability of DSL networks in the presence of slowly varying noise. In this work, we present a novel distributed approach to the margin optimization problem where the network operator specifies the desired ratios of user margins relative to each other and, given this constraint, the Spectrum Balancing algorithm seeks the solution with the largest sum margin. By employing partial Lagrangian relaxation of a rate-adaptive problem each user optimizes its own transmit power spectrum while the Spectrum Management Center (SMC) iteratively updates the target margins of each user so that the system eventually converges to an optimal power allocation with desired margin ratios. Numerical simulations demonstrate fast convergence and show that significantly higher margins can be achieved with our approach compared to purely selfish optimization.