Research On Congestion Control In Streaming Media Transmission

With the extensive use of streaming media and people's pursuit of the high quality of service, streaming media transmission technology is becoming increasingly important. Streaming media has the characteristics of large amount of data, low delay, and low jitter. In addition, because TCP uses the congestion control mechanism of halving its sending rate when congestion happen, which may cause data flow sending rate volatility, TCP is not suitable for the transmission of streaming media. But using UDP may increase network congestion and TCP flows' "starvation" phenomenon. These rise new requirements for congestion control of the streaming media:on one hand network congestion control is needed to maintain transmission flows and other business friendly by the transmission of streaming media, on the other hand the transmission rate of streaming media should keep stability, the rate fluctuations can not be too severe. In addition, with the diversity of streaming media service levels, network transmission of streaming media is facing new challenges and opportunities.Based on the above requirements, the thesis studies the network performance of stream media, focusing on the end to end and router-based network congestion control mechanism. The main works are as follows:(1) In the end to end congestion control mechanism, based on the features of the rate change of streaming media should not exceed 25%, the avoidance phase of TFRC should be improved, resulting in a new STFRC (Smoothness for TFRC) algorithm that solves well the problem of TCP-friendliness, makes the rate change more smoothly on the TFRC congestion avoidance phase. It is more suitable for streaming media transmission; (2) In router-based congestion control, the analysis result of RIO-C and WRED shows that RIO can guarantee the absolute distinction between different priorities of streaming media, while WRED is to provide relative distinction, so RIO-C is more suitable for the transmission of streaming media. However, further study of RIO-C finds that the RIO-C inherited the parameters sensitive problem of RED. Combined with the idea of ARED, RIO-C is improved resulting in another DRIO-C (Delay for RIO-C) algorithm. Simulation results show that the DRIO-C is more suitable than RIO-C to solve the transmission of streaming media.