Size-based scheduling to improve fairness and performance in 802.11 networks

Using wireless technologies to access the Internet (e.g., public hotspots) and setting small ad hoc networks (e.g., during a meeting), are common experiences for end users. Those environments are challenging for TCP because of the variability of the available bandwidth and the competition among flows that span over paths of different lengths and different types (wired or wireless). In this paper, we advocate that part of the problem comes from the use of the FIFO scheduling policy. We propose to replace FIFO with a size-based scheduling policy called LAS (Least Attained Service). We demonstrate that LAS solves most of the fairness issues encountered in those environments and makes a better use of the scarce bandwidth resources.