Video transport over packet-switched networks
Anujan Varma, Christos Tryfonas · 1999
This dissertation deals with several problems encountered while transporting video traffic over packet-switched networks. The focus is on the transport of MPEG-encoded video streams over ATM networks, although most of the results apply to more general environments. Important issues covered in this dissertation include the clock recovery problem at the receiver for video applications with stringent clock requirements. Since packet delay variation (jitter) is present in a packet-switched network environment, clock recovery is often a non-trivial task. We also consider the traffic characterization problem for video sources to facilitate the call-admission control and the traffic policing processes performed by the packet network. We start by studying the clock recovery problem for MPEG-2 Systems Layer streams as seen at the decoder end. A new architecture for decoder design is proposed which is based on a jitter estimator capable of performing restamping on the incoming packets containing clock values in order to minimize the effects of jitter from sources other than the frequency difference between sender and receiver. Next, we deal with the clock recovery problem as seen at the encoder end. We provide guidelines for selecting the timestamping interval for transmission of the Program Clock Reference (PCR) timestamps in a packetized MPEG-2 transport stream sent using the PCR-unaware encapsulation scheme. We continue with an analysis of the effect of the correlation of the delay experienced by the video traffic stream on the quality of the clock recovered by an MPEG-2 decoder. We prove analytically that, for a general input process, high correlation of the delay samples produces a large variance of the recovered dock. We verify the analytical result through actual simulation of an MPEG-2 Systems decoder PLL with correlated traffic patterns (AR processes) as the input. Finally, we study the problem of characterizing the burstiness and loss behavior of deterministic video sources. Such characterization is essential for determination of the network resources (bandwidth, buffer space, etc.) for supporting the desired level of quality when the video stream is transported over a packet network. We develop efficient deterministic algorithms for computation of the burstiness and loss curves of both elementary video streams and MPEG-2 Transport Streams. The algorithms enable the exact computation of the burstiness and loss curves of a video stream, as compared to simulation-based algorithms where the accuracy depends on the rate granularity chosen. The algorithms exploit the piecewise linearity of the burstiness and loss curves, and compute only the points at which the slope changes. The algorithms may be used for the burstiness- or loss-characterization of any bursty on-off source, including voice and data. The computational efficiencies of the algorithms make them attractive for constructing the burstiness and loss curves in a wide range of applications, ranging from video-on-demand servers to real-time video distribution systems. (Abstract shortened by UMI.)