A mathematical model of westwood+ TCP congestion control algorithm

Classic TCP congestion control implements the Additive-Increase/Multiplicative-Decrease (AIMD) probing paradigm to track the Internet time varying available bandwidth. The recent Westwood+ TCP is a sender-side only enhancement of the classic Tahoe/Reno/NewReno TCP that proposes to estimate the bandwidth available for a TCP connection by properly counting and filtering the stream of ACK packets. The estimate is used to adaptively decrease the congestion window and slow start threshold after congestion. In this way Westwood+ TCP substitutes the classic multiplicative decrease paradigm with an adaptive decrease paradigm. This paper proposes a mathematical analysis of the new Additive-Increase/Adaptive-Decrease (AIAD) paradigm. In particular, three equation models have been derived to predict the steady-state throughput of Westwood+ with different degree of approximations. Based on these models, we have shown: (1) the global and exponential stability of the steady-state equilibrium point; (2) the friendliness of the AIAD towards the AIMD algorithm and (3) the improved fairness provided by the AIAD paradigm w. r. t. AIMD paradigm. Finally, the three throughput equation models have been validated using ns-2 simulations in the presence of a wide range of network loads and loss probabilities. Results have shown that the proposed models predict the throughput of AIAD controlled flows with an average error within 15%.