Research On Congestion Control Stratesies Of Industrial Network Using UDP

In the process of real project development, the congestion will happen for the devices that generating and sending data need to be run continued and uninterrupted when most of data in the industrial network transported by UDP, which leads to the increasing of the amount of data. A slight congestion will cause lost packets increase, and even causes the collapse of the network, apart from this, lots of UDP flow will plunder and occupy the bandwidth of TCP, resulting in the throughput of TCP decline rapidly, so it is necessary to make a congestion control of UDP to prevent the UDP flow unrestricted growth, and ensure the fairness to TCP.This article conducts a deep research on the congestion control strategy based on source system and router, carries on the analysis on the congestion control algorithm from the network source and router. And then on those basis, improve those algorithms.First of all, analyses the TFRC(TCP-Friendly Rate Control) algorithm and then improve the shortcomings of response lag of congestion and lack of stratification, we get MLC-TFRC(Multi-Level Congestion TFRC) algorithm. The algorithm first calculates the delay, then obtain the jitter, and then according to the jitter prejudge whether the congestion will occurs; after that, using the different throughout models at the sending source position to quickly solve the network congestion according to the different level of network congestion.Secondly, elaborate the AQM(Active Queue Management) mechanism, analyses the ECN(Explicit Congestion Notification) algorithm, aiming at the shortcomings of ECN, propose EQCN(Explicit Queue Congestion Notification) algorithm, which based on the explicit queue. First modify IP packet head, then add the EQ flag in the identification area and two options in the option area of IP head which use for save the length of queue and the number of UDP packets. Each round the receiving source calculate the two values, and fill them into the options. In addition, improve the RED(Random Early Detection) to the QRED(Queue-based Random Early Detection) algorithm. The improved algorithm changes the single linear function(drop rate-average length of queue) of RED to multi-linear function which have different slope, which will help to make the drop rate move stable, meet the relevant requirement of network. Combine the EQCN algorithm and QRED algorithm to EQ-RED(Explicit Queue based RED).Finally, simulate the algorithms TFRC and MLC-TFRC with the softwoare NS2(Network Simulator 2), then analysis the simulation results from throughput, loss rate and TCP friendly aspects, prove that the improved algorithms' performance has a significant improvement compared to the previous algorithms, which verifying the validity of the algorithms. Then, analysis EQ-RED from the aspects of average queue length, delay and loss rate, prove the improvement is effective.