Assessing the quality of multimedia communications over Internet backbone networks

The Internet is evolving to become the universal network infrastructure that supports all traffic types and communication needs. Among them, interactive voice and video applications, such as Voice over IP and Video Conferencing, are of great interest. As these applications become increasingly popular, users expect higher quality. However, they both have strict requirements both in terms of loss (for good speech and video quality) and delay (for interactivity). If the Internet is to eventually replace traditional networks, such as the telephone network for voice communications, it has to provide service at the same quality levels. Our objective in this study is to assess to what extent today's Internet meets this expectation. In the process, we identify sources of impairments and we comment on the possible causes and improvements.; We first study loss and delay measurements collected for many days, over the wide-area backbone networks of several service providers across the US. We then study the performance of voice communications over these networks, in terms of user perceived quality. While backbone networks are generally believed to be sufficiently provisioned for data applications, we find that this is not always the case for voice and video traffic. Loss and delay characteristics are not consistent across all backbone networks. Some backbone networks exhibit good characteristics and may offer communication at high quality levels, leading to a confirmation that packet voice is a sound approach. Other backbone networks exhibit undesirable characteristics, such as high delay jitter, periodic delay patterns and long loss periods. At the receiving end, playout scheduling plays an important role in smoothing out the delay jitter caused by the network. We examine existing playout scheduling algorithms and we propose new algorithms that take into account the delay variability observed in the measurements and the loss-delay tradeoff in the overall VoIP performance. We finally study the performance of video traffic with low delay constraints and we find it to share fate with voice over the same backbone networks.