IJCSNS International Journal of Computer Science and Network Security, VOL.9 No.4, April 2009 62 Manuscript received April 5, 2009 Manuscript revised April 20, 2009 An Effective Mechanism for Congestion Control in High Speed Networks Ms.T.Sheela’ and Dr.J.Raja’’ Sathyabama University, SSN College of Engineering, INDIA Summary Due to the recent trends in Internet, for exchange of information in the form of pure data traffic and multi- media traffic, High-speed network is necessary. As there is a growing demand for high-speed networks, data transfer must take place without any congestion. In data networking and queueing theory, Network congestion occurs when a link or node is carrying so much data that its quality of service deteriorates. Typical effects include queueing delay, packet loss or the blocking of new connections. A consequence of these latter two is that incremental increases in offered load lead either only to small increase in network throughput, or to an actual reduction in network throughput. The Transmission Control Protocol (TCP) is one of the core protocols of the Internet Protocol Suite. TCP has not performed well on high-speed network because the standard TCP’s algorithm for congestion control may cause thousands of packet drops in one Round Trip Time (RTT) and the window size is halved at the time of congestion. So the utilization of the bandwidth and throughput is minimized. The researchers developed different TCP variants to improve the performance of the congestion control algorithms in high-speed network. In this paper we propose a sequence of algorithms based on window adjustment, feedback mechanism and buffer management to overcome the limitations for high-speed network. The congestion control mechanism that we adopt starts with Queue Management, which enables to modify and control the transmission queue. The Window Adjustment Mechanism alters the Congestion window size based on the Input load. The Feedback mechanism renders the optimum load that can be serviced currently when congestion had taken place. Also the feedback information is notified to the sender. This solution to control the congestion in the network achieves maximum throughput and bandwidth utilization, with minimum delay and drop probability. The performance of the congestion window, throughput, utilization of the bandwidth, delay are analyzed and presented. Keywords: TCP/IP, Congestion Control, Window Management, Feedback Control, Queue Management, Explicit Control. 1. INTRODUCTION As with increase in the amount of data transfer across the various networks, to achieve low delay, maximum throughput and predictable performance on an end-to-end basis[1][5][9], a high bandwidth environment is required. An increase in bandwidth and data rate leads to congestion. So congestion has to be avoided while sending large volumes of data within short period of time. Congestion window mechanism in TCP increases the window size to their threshold value and drastically decreases at the time of congestion in high-speed network[3][6][12]. This is because TCP increases and decreases its congestion window size too slowly with non- duplicate ACK packets [10][13][18]. If we require responses to be quicker, then a connection with large RTT must be present than a connection with smaller RTT, because increase of TCP is tied to each RTT. If large amounts of data have to be sent then TCP sends all these packets based on the window size and which can increase the congestion problems, because the window size is halved [25]. In TCP, loss of packets by itself acts as a signal which indicates the sender to lower down the congestion window to limit the number of packets to be sent and thereby congestion can be reduced. Thus TCP initiates congestion control only if a packet is considered lost. To recover from the packet loss, congestion window size is reduced by half, and so the congestion window size is increased by 1 segment for every Round Trip Time. For example, in 10 Gbps connection, initially the link will operate at 10 bps connections and then gradually increase to 10 Gbps and hence by means of TCP it will take more time to transfer data on high-speed networks[12][13]. Due to the gradual increase and sudden decrease in high-speed network, utilization of bandwidth by TCP is very poor . Another issue with TCP is the way it allocate bandwidth on networks with high bandwidth delay products and also the Queue size [8][14]. If we have a fast link (eg. Gbps), the TCP’s AIMD algorithm is not able to send large volume of data because the sending rate is very low. So automatically limited number of packets are