Improving performance of adhoc TCP in Mobile Adhoc Networks

Modifying transmission control protocol (TCP) to improve its performance in wireless networks has been a long-standing research problem. Many methods have been proposed to improve TCPpsilas performance in mobile adhoc networks (MANETs). Among them ad hoc TCP (ADTCP) uses an end-to-end approach which requires minimal changes at the sender and receiver, provides the flexibility for backward compatibility, maintains end-to-end TCP semantics and is TCP-friendly. It uses end-to-end measurements to detect congestion, disconnection, route change, and channel error, and each detection result triggers corresponding control actions. This paper proposes improved-ADTCP (I-ADTCP) an improvement on ADTCP. To improve the performance of ADTCP we consider following: ensure sufficient bandwidth utilization of the sender-receiver path; avoid the overloading of network by limiting TCPpsilas congestion window below the upper bound of bandwidth delay product (BDP-UB) of the path; check for incipient congestion by calculating inter-packet delay difference and short term throughput using relative sample density (RSD) technique. In an incipient congestion condition, we reduce congestion window limit (CWL) to half, which limits the packets send by the sender and does not allow the congestion to build up. Thus the algorithm tries to remain in congestion avoidance phase at all times by detecting and reacting to incipient congestion. Results of simulation using NS-2 show that I-ADTCP performance is superior to that of ADTCP for all levels of traffic intensity in the network. The good put performance is improved by 10%-30% in I-ADTCP.