Composite TCP: Refining Congestion Control

Internet users need reliable transmissions for web browsing, email, file transfer, and database access. TCP is the dominant reliable transport protocol on top of which all of these services run. Since TCP is originally developed to be used on the wired network, it safely assumes that segment losses are due to congestion. This is not true for wireless media where, due to fading channels and user mobility, transmission errors are more frequent. It unconditionally reduces its flow when a packet loss is detected, assuming that it has occurred because of congestion in the network. This incorrect assumption in case of a packet loss occurring due to channel noise, adversely affects the performance of TCP. This paper focuses on the same problem of TCP related to random errors causing packet loss. Treatment for losses should be different for corruption and congestion to improve efficiency. The proposed Composite TCP (CTCP) utilizes the SACK information available from a receiver along with round trip time to avoid flow control when the loss is because of a random error. The performance b of Composite TCP is compared and analyzed with existing variants over networks with exhaustive simulations using ns-2. Simulation results indicate that the proposed protocol significantly improves performance.