Comparison of VBR and CBR service classes.pdf

Comparison of VBR and CBR service classes for MPEG-2 video multiplexing - Is statistical multiplexing really useful for bursty video transmission? - Naoki Wakamiya, Masayuki Murata, Hideo Miyahara Graduate School of Engineering Science, Osaka University 1-3 Machikaneyama, Toyonaka, Osaka 560-8531, Japan SUMMARY It has been believed that a statistical multiplexing technique can bring an effective bandwidth usage. When it is applied to MPEG-2 video, it has been pointed out that two to several times multiplexing gain can be attained due to highly bursty nature of video traffic. However, most of past researches do not take into account the implementation aspects of this technique. In this paper, we thoroughly investigate the actual multiplexing gain by comparing ATM CBR and VBR service classes (determin- istic and statistical service classes), and point out that the sta- tistical multiplexing gain is far from one may expect. Keywords: MPEG-2, ATM, VBR service, CBR service, statistical multiplexing gain I. INTRODUCTION There is an increasing interest in the network support of the effective and high-quality video transfer. Because it is very ex- pensive to transfer un-compressed videos, several coding and compression algorithms are developed and used, such as mo- tion JPEG [1], H.261 [2], MPEG-1 [3] and MPEG-2 [4]. These coding algorithms are different from each other, but gener- ated coded video traffics have the same characteristics called “burstiness”. Burstiness means that the traffic rate fluctuates depending on time, and is usually measured by peak to aver- age ratio of the traffic rate. For example, the peak rate becomes two, three, or, sometimes several times higher than the average rate when a video stream is compressed with the MPEG-2 cod- ing algorithm. This highly bursty nature of coded video traffic seems to be well suited to the statistical multiplexing strategy of the high speed networks such as ATM (Asynchronous Transfer Mode) networks. However, multimedia