Improving And Simulating Of Slow-Start For TCP Congestion Control Schemes

TCP congestion control algorithm was consisted of about four kinds of algorithms. They are as follows. 1. Slow-start, when setting up connection or connection halted meanwhile waiting for over-time occurring, injects the data stream into network using slow-start strategy, and gradually tests out the available bandwidth. 2. Congestion avoidance, forecasting when the congestion will be happen, and then reduce the transmission rate to the host computer before the group just will be discard, so that to avoid network congestion. 3. Fast retransmit and fast recovery, transmit the dropped data groups again and come back from the congestion, when there are data groups being dropped or the retransmission is overtime. The underlying idea of the algorithm described above is the AIMD (Additive Increase and Multiplicative Decrease) proposed by Jacobson in 1988. At that time, the data flow in the internet was less and, while nowadays, it is increasing rapidly and the kind of it is various, too. According to the time that the connection of data transmission costs, we can sort these connections into two classes: long term linkages and short-term linkages. Researches show that most of the linkages that transmit the data groups in the Internet are short-term linkages, such as WEB, but the long-term linkages transmit the most data stream such as FTP. Short-term linkages always does not level off but terminate, it means the short term linkages is often on ground of slow start phase. So we can conclude that the performance of the slow -start directly affect the transmit efficiency of the short term linkages. For the long-term linkages the slow start will play the role as follows: start-up the connection and the retransmit the groups, so that the performance of the slow-start also influences the long-term linkages. Also we can say advancing the performance of the slow-start is useful for the performance of the Internet.This paper aims at improving the performance of the slow-start algorithm primarily. In the former algorithm, the datagram increases too fast, and it would be rise the probability of congestion. The paper enhances the former method for the problem by reducing the increasing speed of the datagram, and thus decreases the occurring rate of congestion and improves the performance of network. The probability that data stream appear suddenly will be reduce, so that the congestion can be decrease, and the performance of the network will be improved. Using the network simulating NS-2 to check up the algorithm, the experiment shows that the algorithm can reduce the risk of network congestion, and improve the transmission performance for network.