Achieving per-flow fair rate allocation within Diffserv

This paper addresses the fundamental issue of providing per-flow fairness within the Diffserv framework. The fair allocation derivative estimation (FADE) algorithm for estimating flow fair share in the absence of per-flow information is proposed. FADE calculates fair share feedback using a modified quasi-Newton method. This efficient method for estimating fair share provides a more precise model than other existing fairness estimation approaches. As such, it is able to more accurately estimate fair share and quickly converge to the proper rate. The simulation compares FADE to other proposals and demonstrates the overall effectiveness of the algorithm.