Optimizing TCP Vegas for Optical Networks: a Fuzzy Logic Approach

Performance of TCP is reduced over buffer-less optical burst switched (OBS) networks by misunderstanding of the congestion status in the network. In other words, when a burst drop occurs in the network and we cannot distinguish congestion and burst contention in the network, TCP wrongly decreases the congestion window size (cwnd) and causes significant reduction of the network performance. This paper employs the fuzzy logic to solve this problem. By using the fuzzy logic we provide a framework to distinguish whether the burst drop is due to the congestion or is due to the burst contention. The full approach, for detecting state of network, relies on Round-Trip-Time (RTT) measurement only. So, this is an end-to-end scheme which only end nodes are needed to cooperate. Extensive simulative studies show that the proposed algorithm outperforms other TCP flavors such as TCP Vegas, TCP Sack and TCP Reno, in terms of throughput, packet delivery count and fairness.