Fuzzy Explicit Window Adaptation: a Method to Further Enhance TCP Performance

In the current Internet, standard TCP congestion control is based on segment loss events possibly assisted by simple explicit router feedback mechanisms like Explicit Congestion Notification (ECN). The performance of TCP congestion control can be improved if a more sophisticated Router Congestion Feedback (RCF) mechanism than ECN is used. One of such approaches is Explicit Window Adaptation (EWA). EWA uses TCP’s built-in flow control mechanism to inform TCP senders about the current load in a network path. To reach this, an EWA-capable router is able to decrease the advertised receiver window in TCP acknowledgments if necessary in order to avoid congestion and packet losses. Since the EWA algorithm performed in a router has some shortcomings under certain conditions, a new EWA-related approach is developed. This approach, called fuzzy explicit window adaptation (FEWA), replaces parts of the EWA algorithm in a router with a more accurate fuzzy-based calculation. In this technical report, FEWA is described in detail. Using two simulation scenarios with different load in the network, the performance of standard TCP is compared with the performances of standard TCP assisted by EWA and standard TCP assisted by FEWA. The simulation results show that standard TCP assisted by FEWA outperform both other approaches in throughput over loaded routers and considerably decreases the number of packet losses in congested routers.