The Research Of The Improved TCP Congestion Control Algorithm

The increase of Internet user and the data transferred in the Internet result in more and more serious congestion. The goal of congestion control is to ensure gaining certain throughout to increase the use of bandwidth, in the mean time, to avoid congestion collapse by taking certain control mechanism. The key of the Internet stability and robustness is congestion control. No matter what control mechanism is adopted, to control congestion efficiently, an affective algorithm is needed.There are two congestion control algorithms, one is source algorithm executed by host computers and edges, and the other is link algorithm executed by network devices. The basic idea of source algorithm is to adjust dynamically the sending rate or the size of the congestion window. Source arithmetic includes four stages, they are slow start, congestion avoidance, fast retransmit, and fast recovery. The kernel of the source algorithm is AIMD (additive increase multiplicative decrease).Along with the update of network devices and the advance of network technology, Internet service provider can provide wider bandwidth, hi a wide network, to keep a higher throughout, source must hold a bigger sending window. But when the congestion takes place, using the traditional TCP congestion control algorithm, the network needs longer recovery time, and the sending window will be changed from a high size to a low one, it will result the rapid vibration. The bandwidth can not be made use of and the network resource will be wasted in vain.Aim at the weakness of the traditional TCP congestion control algorithm, the paper proposes a novel congestion control algorithm, the key idea of the algorithm is: Initially:cwnd = 1;ssthresh = infinite (64K);For each newly ACKed segment:if (cwnd < ssthresh) /* slow start*/ cwnd = cwnd + 1;else/* congestion avoidance;cwnd increases by 1 per RTT */ cwnd += 1/cwnd;Triple-duplicate ACKs:/* multiplicative decrease */ cwnd = ssthresh = cwnd/2;Timeout:ssthresh=cwnd-cwnd/8;cwnd= max(2, min(cwnd/2, awin));cwnd: congestion windowawin: the flow control window that controls the sending rate of the sending-end to avoid the receiving-end cannot receive data comes from source, ssthresh: slow start thresholdThe analysis of NS-2 (network simulator version 2) result proves that the novel congestion control algorithm can speed congestion recovery on a certain extent, and decrease the vibration, and economize the bandwidth accordingly. In addition, in the process of NS-2 result analysis, a new method be found, the NS-2 simulating result file out.tr (text file) can be become a DBF-type file, with the DBF file, it is very convenient to gain correlative data and statistical results by individual program. By the method, more precise and useful information can be obtained from the DBF file.The NS-2 experiment shows that the novel congestion control algorithm wastes little recovery time than traditional TCP congestion control algorithm, the novel congestion control algorithm has the advantage of good stability, so the network can be made use of by a long way. FCA may be a reference in IPV6.