ABSTRACT
Congestion in the network impacts video transport in all three dimensions: inconsistent bandwidth availability for the video stream, increased latency and jitter, and packet loss. Even when video streams have priority for transport over the network, potential sources of congestion in the network may be other video streams competing for capacity, or other even higher priority flows, such as net work control traffic. Because of the high variability of the bandwidth require ments even for a single video source (see, e.g., Ref. 1), congestion due to competing video sources is likely to pose a major challenge to maintaining qual ity. The first defense against congestion is to utilize buffers in the network to
overcome short-term mismatches between the available capacity in the network and the rate required for transporting the video stream. However, when the rate mismatch lasts a sufficiently long period (maybe a few milliseconds, or some times even less, depending on the amount of buffering available), packet loss re sults. Since video is typically transported over the network using an unreliable transport protocol such as UDP, packet losses typically result in the loss of video frames. Depending on the encoding, video has varying amounts of tolerance to frame loss. Typically, no more than a very small probability of frame loss is ac ceptable. Frame loss arises from the loss of one or more packets, depending on the encoding. The requirement for video on the network transport is to have a very small packet loss probability, in the range of 0.1-5%, depending on use. This is quite difficult to achieve, unless the network is considerably overprovi sioned.
