In general a retransmission timeout (RTO) results from the congestion in the network and reduces TCP flow drastically. TCP may experience RTO even in absence of congestion particularly due to loss of retransmission. In this paper, a simple technique is proposed to identify those timeouts, which are not due to congestion and should be avoided. These avoidable timeouts are typically caused by loss of retransmission due to random channel errors. This type of retransmission timeouts is common in wireless environments in general and satellite links in particular. A TCP sender refrains from transmitting packets for a considerable time i.e., the timeout period. This paper proposes a modification in the original SACK TCP to identify the avoidable timeouts and also to utilize available RTTs before such timeouts. Improved TCP performance of the proposed algorithm to combat channel noise during retransmissions is substantiated with the help of simulations using open source software (ns-2). The simulation approach and its results are also analyzed in this paper.

Deployment of wireless links (terrestrial and satellite) along with wired links has made extension of the Internet even in remote places feasible. TCP/IP protocol suite is an integral part of the Internet. Congestion control of TCP plays a vital role in the performance of the Internet. TCP's unconditional flow control in case of a packet loss has always been a concern for researchers. Further, halving congestion window in such conditions without taking in to account the current network state is also considered inappropriate. The problem is compounded in wireless networks where packet losses occur often due to channel errors rather than the shortfall in the available bandwidth. In this situation, TCP's conservative behaviour underutilises the bandwidth. We therefore, propose a scheme to address the issue of underutilization of network resources. The proposed approach, Discrete TCP (DTCP), differentiates slow start and congestion avoidance phases while tuning data flow over a transport connection. DTCP evaluates ssthresh and cwnd before setting up parameters, based on the existing network condition to enhance the performance. The proposed scheme is compared and analyzed with various existing schemes with the help of extensive simulations using ns2. Results of simulation based experiments indicate significant performance improvement of DTCP on erroneous links and in heterogeneous networks and confirm its suitability.

No abstract

No abstract

No abstract