High-speed network support for multimedia traffic
Rose P. Tsang · 1996
The deployment of high speed network will enables the development of a plethora of new distributed multimedia applications. These applications are based upon the transmission of digital information such as video, voice, high-resolution images, graphics as well as traditional computer data traffic. The challenge lies in efficiently integrating these various data types onto a single communications infrastructure which common switching and transmission facilities. Continuous media, such as digital video, is particularly interesting due to its high demands in terms of bandwidth, real-time delivery and processing requirements. Experimental results of transmitting digital coded video over several Asynchronous Transfer Mode network platforms are presented. The impact of high speed network components' characteristics, such as switch fabric architecture, buffering strategies, buffer dimensioning and traffic control, are illustrated through the experimental results. Traffic control for continuos media traffic is an inherently difficult problem due to the time-sensitive nature of the traffic and its unpredictable burstiness. We present several algorithms which control traffic by dynamically allocating/de-allocating resources among competing connections based upon their real-time requirements. These schemes incorporate rate-control, buffer control, real-time burst-level scheduling and link-link flow control. We show analytically potential performance improvements of our traffic control methods and present a scheme for buffer dimensioning. We also present simulation results and discuss the tradeoffs inherent in maintaining high network utilization and statistically guaranteeing many users' Quality of service. Another important characteristic of continuous media traffic is its connection-oriented nature. We present a study of efficient arrangements of connection-oriented switches. Given a number of station nodes attached to a set of switches, minimal hardware requirements as well as the interconnections topology which supports the maximal aggregate network throughput are sought. Two classes of network topologies based upon switch usage and station node distribution are presented and analyzed. One of the classes is shown to provide optimal network throughput (i.e., maximal number of simultaneous connections). the hardware requirements to ensure nonblocking for this topology are derived. Blocking probabilities for blocking versions of this topology are also derived.