Stability and fairness of rate estimation-based AIAD congestion control in TCP

We analyze two achievable rate estimators that use different timestamps of consecutive packets. We examine the effect of the choice of rate estimators on the stability and fairness of a class of TCP protocols that use this estimated rate to implement the additive-increase/adaptive-decrease (AIAD) congestion control. Simulation results confirm our analysis that rate estimation based on the inter-arrival times of the ACK packets is not properly bounded and would cause instability of the AIAD algorithm and unfairness among competing TCP flows, particularly in networks with small or moderate buffer space. Whereas, the rate estimation based on the inter-arrival times of the data packet at the receiver maintains its accuracy even when the reverse path is congested and enables the AIAD algorithm to maintain the stability and fairness as the number of competing flows increases. Our analysis also suggests a straightforward enhancement to TCP Westwood that would improve its stability and fairness. The enhanced algorithm can be easily implemented without any modifications to the TCP receiver-side code by enabling the TCP timestamps option.