Publication

Abstract : In Ad-hoc network, Packet reorder is effect of many causes like congestion loss, packet loss due to corruption, multi-path routing and node's mobility. Transmission Control Protocol (TCP) performs poorly over wireless links which misinterprets the reordered packets as lost packets due to congestion. This has motivated researchers on developing TCP variants towards packet reordering problem for enhancing TCP performance. Some researchers have found the opportunity to use existing TCP variants like SACK, TCP-Westwood to address the reordering problem. TCP-DCR (TCP-Delayed Congestion Response) and RR-TCP (Reordering Robust TCP) are the TCP variants to acknowledge the packet reordering at the end level. In the existing investigations, the protocols were tested in a simulation environment with one or two causes (congestion or error) at a time over a simple topology. In our experiments, the simulation study of TCP variants has been done with all reordering causes simultaneously, over a large ad-hoc environment including mobility and multi-path routing. A performance analysis had been done using reordering metric (percentage of reordering) which accounts for earlier and lately arriving reordered packets and estimated using throughput and goodput metrics. From our analysis, it is observed that the Selective acknowledgement option of SACK-TCP and the delayed congestion response of DCR help to achieve better performance than other variants in presence of low reordering due to high congestion and low error. However, these variants were failed to meet up their performance in case of high reordering (reordering greater than 50%) due to high congestion loss and high channel error losses. This observation has given the research opportunity to incorporate a discrimination algorithm in the existing reordering solutions to classify the packet error from packet reordering.