Internet Traffic And Congestion Control Problems

As is known to all, Internet is mainly based on TCP/IP protocol, which provides only best effort service. With the development of Internet and E-commerce, there is an increasing demand on the improvement of Internet traffic service. A good way to better this service is to improve its algorithms regarding flow regulation and congestion control. This paper tries to solve the problem in two ways. First, in order to improve the network resource allocation among TCP flows, an ideal resource allocation model is put forward and implemented by flow marking and K mean clustering methods. Second, in order to effectively control the unresponsive flows that cause network congestion, a simple and feasible method is brought up. It consists of two major parts: domain based feedback mechanism and drop driven AIMD simulation. In comparison with others, this method has a limited upper bound of communication cost, with no need of dynamic calculation of RTT and consumption of extra buffer from routers. These features make it applicable in a broad range of areas.