Reasearch On The Multimedia Traffic Congestion Control Algorithm In Hosts And PID Algorithm In Routers

With the evolvement of Internet based on TCP/IP, the scale, users and trafficshave increased rapidly since 1990's. The network congestion has become moreserious and complex due to the ever-increasing network application types anddynamic network parameters such as active sessions and round trip time.Congestion often results in decline of quality of service (QoS) in terms oftransmission delay and throughput, while the network resource utilization likebandwidth and buffers are also affected seriously. The congestion control is alwaysa hot spot in the field of network research. Currently, there are two kinds of theschemes of network's congestion control. The first is congestion control strategiesin subnet. The second is congestion control strategies in hosts.The congestion control strategies in subnet are set in the intermediate nodessuch as router, and work on the network layer of the OSI model. It can be classifiedinto queue management algorithm and queue scheduling algorithm, of which thequeue management algorithm is the main field. The queue scheduling algorithmallocates the bandwidth through the rules of schedule. The queue managementalgorithm allocates the buffer by controlling the drop rate decided by the length ofthe queue in the buffer.The way of scheduling the traffic can be divided into one-queue andmulti-queue. When one-queue is used, the received data are put into one queuewithout different, and are transported in the way of FIFO. When multi-queue isused, the received data are put into several queues with the rules. The factorsconsidered to estimate the algorithm include delay, friendliness, complexity etc.In order to control the congestion, the queue management algorithm maintainsthe length of the queue by using a special rule dropping some of the packetsreceived. When the packet is dropped, the host will be noticed and decline thesending rate. There are two sorts of queue management algorithm, one is thepassive queue management and the other is the active queue management.DropTail is the primary passive queue management;it is used in the early network.With the development of the network, the subnet is needed to participate incontrolling. So the active queue management algorithm is advanced, which hasbecome the main spot of the congestion control research.C. Hollot inducted classical control theory into the designing of AQMalgorithm, and designed PI controller using frequency-correctting method in 2001.Because PI controller is not accurate on choosing parameters with tring means andits transient-state performance is not good enough. Ren Feng-yan designed PIDcontroller. The differential link was inducted into PI Controller to enhanceresponse performance of the systems, and proposed the method of choosingparameters based on stabilization balance. These make a good guarantee to thestability of the controller. Classical PID controller has simple structure, it is easy toadjust, and the parameter choosing process is closed to engineering objective. Butit is hard to adjust parameters on-line. It is difficult to control load process and slowtime-varying system effectively.The congestion control strategies in hosts tackle the congestion from thetransmission of the traffic. In order to supply the good quality service, it controlsthe ways of the hosts sending the data to adjust the situation of the intermediatenodes. According to the OSI model, the congestion control strategies in hosts workon the transport layer and higher layer. Currently, the research of the congestioncontrol strategies in hosts concentrates on the TCP congestion control andTCP-friendly congestion control.Internet primarily relies on TCP end-to-end congestion control to avoidnetwork congestion. But TCP strategies couldn't meet many various demands ofevery application on the complex network. In fact, it will be more effective fordetecting and preventing congestion if the routers perform queue managementschemes. Congestion control strategies for the routers have been discussed widely.Multimedia applications such as video and audio have come forth continuously inrecent years. Multimedia traffic is sencitive to delay, and needs stable sending rate.It is necessary for multimedia's sending rate to avoid changing acutely. Howeverwhile TCP congestion control is appropriate for applications such as bulk datatransfer, many real-time applications would find halving the sending rate of a flowto be too severe a response to a congestion indication, as it can noticeably reducethe flow's user-perceived quality. UDP protocol is employed to transfer multimediaflows. However, UDP flows are not friendly to TCP flows because they have nocongestion control scheme.Congestion control strategies for IP based networks are introduced in detailfrom two different aspects, layers of routers and end host. Systematic research tomeet practical requirements of IP based networks is performed in depth. Theprimary works and innovations of this paper include but not limited to:(1) Aim at the disadvantage of on-line adujustment and the weak adaptiveability to the normal PID controller parameter, the paper designs the on-line neuraladaptive PID controller for queue management in networks, which can correct thecontroller parameters continuously by adjusting the system errors, can adapt itselfto the parameter changes of controlled structure or network environment, andstabilize the queue length of the router. The results of simulation show that thepresented adaptive neuron PID controller can effectively avoid network congestion,and stabilize queue length to the desire value fast. The algorithm keeps the higherlink utilization rate, and has a good robustness.(2) The congestion control mechanism of TCP is not very suitable formultimedia traffics because of its characteristic. When a packet is discarded, thesender halves its rate, which results in sending rate wobbling greatly. Many ofthese new applications use UDP, but UDP has no congestion control mechanism,which may harm the robustness of the Internet and eventually lead to congestioncollapse. This paper presents an enhanced congestion control algorithm based onAdditive Increase Multiplicative Decrease (AIMD), which is called as SlowIncrease Slow Decrease (SISD).Since the dominant Internet traffic is TCP-based, it is important that newcongestion control schemes be TCP-friendly. By this, we mean that the sendingrate of a non-TCP flow should be approximately the same as that of a TCP flowunder the same conditions of round-trip time and packet loss. We study the sendingrate as a steady state metric, and responsiveness, aggressiveness and ratefluctuation as transient metrics in the simulation tests.The simulation results show that SISD algorithm has better smoothness ofsending rate compared with the congestion control algorithm of TCP, whichreduces the fluctuation of sending data and adds the congestion control tomultimedia traffics. This algorithm can act as studied basis of multimedia traffics,and can improve the transmission quality of TCP traffics.