Impact of ATM Traffic Shaping on MPEG-2 Video Quality

This paper presents the impact of traffic shaping on the quality of MPEG-2 video transmissions over ATM networks. The performance evaluation is accomplished by simulating an ATM network using real MPEG2 video streams from video broadcast applications. First, it is shown how cell loss influences the perceived video quality. Cell loss can result in either frame corruption or frame loss, which causes differing video quality impairments. Next, a simulation study is conducted showing how leaky-bucket traffic-shaping influences video quality. It is discussed how shaper parameters have to be chosen to obtain highest video quality. In general, underdimensioning as well as overdimensioning of the shaper should be avoided because it results in video quality degradation. Also, it is shown that the optimal choice of the shaper’s token buffer size depends on the overall motion/burstiness in the video. A well-dimensioned shaper is able to increase video quality significantly as compared to not shaping at all. Most of the research on traffic shaping centers on the characteristic of the shaped traffic rather than on the resulting application performance. This study clearly shows that traffic shaping can actually enhance video quality significantly while it shapes the traffic to adhere to the negotiated QoS (Quality of Service) parameters.