Video streaming from a wired network to a wireless network involves many challenges. The first question is the rate at which a fixed host should stream video to a mobile host. The rate must be optimal in two ways. First, the rate should be fair to other flows (e.g., TCP flows) in a wired network, which has been recognized as TCP-friendly property. On the other hand, the rate should use the wireless bandwidth efficiently. In order to achieve these two goals, the endhost (i.e., sender or receiver) should apply a congestion control mechanism only when there is congestion. The problem then translates into congestion detection in the presense of a wireless channel error. This problem is not easy to solve since both congestion and wireless channel error produce packet loss, which is traditionally a sign of congestion in the Internet.
This problem is well recognized in a wireless TCP senario, and there are some excellent not-end-to-end solutions toward the problem [1,2]. However, these solutions break end-to-end property, and thus introduce scalability problems. Further thinking reveals that those solutions fail when there are asymmetric routes between the sender and receivers. In summary, this problem is not well solved in wireless TCP.
In this project, we will try to explore the possibility of detecting congestion in the presence of a wireless channel error using active UDP probing. This solution is different from many previous solutions in that it is based on pure end-to-end observation. Although it is hard to have an end-to-end solution for TCP, we think it might be possible in a UDP senario, since it is possible to fully control the sending pattern while doing active probing.