1 3 rd IEEE International Advance Computing Conference(IACC-2013)February 22-23, 2013, Ghaziabad (UP)-201009, India Optimistic congestion control to improve the performance of Mobile Ad hoc Network Chandra Prakash Sahu Mahamaya Technical University, Noida, India Prem Shankar Yadav Kanpur Institute of Technology, Kanpur, India Sachin Ahuja Monad University, Hapur, India Rajesh Prasad Ajay Kumar Garg Engineering College, Ghaziabad, India Abstract— TCP Vegas for Ad hoc network is an end to end congestion avoidance protocol that uses conservative approach to determine and control network state. But this conservation scheme is not good in all conditions and it can unnecessarily reduce the size of congestion window. This paper proposes improved TCP Vegas an improvement over TCP Vegas in Ad hoc network that utilizes round trip time variation of packets at sender side, short term throughput and inter-delay difference at receiver side to measure the network state and then controls congestion window considering the path length and network state. Simulation results show the improvement of 5 to 15 % over Ad hoc TCP Vegas in high mobility and high traffic conditions. Keywords—Ad hoc network, Congestion, TCP, Fuzzy Logic, MANET. I. INTRODUCTION As the name suggests the ad hoc network is a temporary arrangement of devices to make communication among them possible. The ad hoc network can be called as infrastructure less network that requires no existing infrastructure, no router or switches, the nodes themselves work as router to provide connectivity to other nodes and as computer for its own [10]. The networks with highly mobile nodes have produced new challenges by introducing some considerations stemming from the special characteristics of the wireless medium and the dynamic nature of the network topology. It requires a robust and highly effective algorithm that can handle the challenges of these networks easily. In MANETs the general TCP does not show very good performance as it fails to tackle the mobility induced through problems like disconnection and reconnection. The route change induces out-of-order delivery of packets at end hosts. The communication is affected by mainly the mobility of the nodes and also the error and contention prone wireless transmissions [3, 4]. Modifying TCP to improve its performance in wireless networks has been a research issue. Many methods have been proposed to improve the performance of TCP in MANETs. Among them TCP Vegas uses an end-to-end approach which is easy to implement and deploy since it requires less changes at end hosts, provides the flexibility for backward compatibility, maintains end-to-end TCP semantics. In this paper we propose an algorithm: improved-TCP Vegas (ITCP Vegas) to improve the performance of Ad hoc network. For this purpose we tune the congestion window size considering network status adaptively by detecting and reacting to different network conditions. The network condition is determined at both end and used for cross verification to reduce the probability of unnecessary reduction in the congestion window size. Since there may be a chance of backward path congestion which can affect the round trip time and thus the decision to control congestion window size. This approach clearly distinguishes forward path congestion from backward path congestion and thus reduces chance of unnecessary reduction of congestion window size. The changes introduced are simple with respect to the implementation complexity. The paper is organized as follows: Section 2 describes the related work in this area, section 3 discusses the proposed approach, section 4 covers proposed algorithm, and section 5 evaluates the proposed algorithm. Finally, Section 6 gives conclusion. II. RELATED WORK In [1], Chen Y., Xue and K. Nahrstedt proposed an adaptive CWL (Congestion Window Limit), measured in number of bytes, setting strategy to dynamically adjust CWL according to the current RTHC (Round Trip Hop Count i.e. the number of hops of the path). To fully utilize the capacity of a network, the TCP connection should set its CWL equal to the BDP (Bandwidth Delay Product i.e. the number of bits that can fill the link)[9]. The bandwidth of the path depends upon the link capacity and the processing capability of nodes in the path. So the path’s bandwidth is taken as the bandwidth of the lowest bandwidth link otherwise the extra bits will cause congestion at the node with minimum available bandwidth. So to utilize available bandwidth the maximum burst size can be taken as the bottleneck bandwidth.