In a mobile ad hoc network (MANET), when a node fails during the RTS-CTS
INTRODUCTIONA MANET is a collection of autonomous mobile wireless nodes which are free to move about arbitrarily. It may operate in isolation, or may have gateways to and interface with a fixed network. The nodes are capable of singlehop or multi-hop connectivity. One important characteristic of MANET is that the network topology may change randomly and rapidly at unpredictable times as the nodes move or adjust their transmission and reception parameters [1]. The network is important in situations where temporary networks are needed but fixed infrastructure does not exist. Some of these applications such as military or serving disaster relief efforts can be time-critical.A MANET is a potential solution whenever a temporary network is needed and no fixed infrastructure exists. Moreover many applications in this environment are time-critical. It is observed that TCP performs poorly in MANET as demonstrated in [2-6, 7, 8]. Here, a node drops a packet either due to congestion in the shared medium or when it is unable to forward the packet to the next hop. A node cannot reach the next hop node because the next hop node has moved out of transmission range or there are simply too many nodes trying to access the channel at the same time [6]. In a MANET, due to the mobility, bandwidth and energy limitations of nodes, link disconnection and packet loss occur frequently. A high level of packet losses and consequently a high number of TCP retransmission time-outs degraded the overall network performance. It should be noted that in MANET, mobility alone can cause a degradation of performance of TCP in ad hoc networks even when the load is light. The objective in this paper is to address the issue of degradation in TCP performance due to mobility.Towards this objective, we propose mechanisms to reduce the number of packet losses and to reduce control/management packet overhead. We are making use of the signal strength measurements at the physical layer to predict possible link failure to a neighbour that is about to move out of range as proposed by [9]. Thus, if the measurements indicate that the link is likely to break, fewer attempts should be tried to restore the link and a search for new route can be initiated more quickly. This reduces packet loss. On the other hand, if the link is very likely to remain within range, the number of retries to restore the link should be increased as to avoid rerouting because rerouting in MANET consumes a lot of network computation power and resources. This reduces control/management packets overhead. Section 2 describes related works. Section 3 and 4 presents the proposed mechanism and its simulation results. Finally, Section 5 concludes the paper with future work.