Dual-Mode Congestion Control Mechanism for Video Services

Abstract

Recent studies have shown that video services represent over half of Internet traffic, with a growing trend. Therefore, video traffic plays a major role in network congestion. Currently on the Internet, congestion control is mainly implemented through overprovisioning and TCP congestion control. Although some video services use TCP to implement their transport services in a manner that actually works, TCP is not an ideal protocol for use by all video applications. For example, UDP is often considered to be more suitable for use by real-time video applications. Unfortunately, UDP does not implement congestion control. Therefore, these UDP-based video services operate without any kind of congestion control support unless congestion control is implemented on the application layer. There are also arguments against massive overprovisioning. Due to these factors, there is still a need to equip video services with proper congestion control. Most of the congestion control mechanisms developed for the use of video services can only offer either low priority or TCP-friendly real-time services. There is no single congestion control mechanism currently that is suitable and can be widely used for all kinds of video services. This thesis provides a study in which a new dual-mode congestion control mechanism is proposed. This mechanism can offer congestion control services for both service types. The mechanism includes two modes, a backward-loading mode and a real-time mode. The backward-loading mode works like a low-priority service where the bandwidth is given away to other connections once the load level of a network is high enough. In contrast, the real-time mode always demands its fair share of the bandwidth. The behavior of the new mechanism and its friendliness toward itself, and the TCP protocol, have been investigated by means of simulations and real network tests. It was found that this kind of congestion control approach could be suitable for video services. The new mechanism worked acceptably. In particular, the mechanism behaved toward itself in a very friendly way in most cases. The averaged TCP fairness was at a good level. In the worst cases, the faster connections received about 1.6 times as much bandwidth as the slower connections. iii Preface This thesis contains the results of research work, which has been carried out at the Pori Campus of the Tampere University of Technology. This work has been a challenging journey, including both uphill and downhill phases. Fortunately, I was not alone on this road but accompanied by a group of people always willing to coach, support, help, and motivate me. Certainly, I would have never reached the endpoint of this trip without the help and support of others. For this reason, I would like to thank all the people who have been involved in the process of creating this thesis throughout the years. First, I would like to thank my supervisors, Professors Pekka Loula and Tarmo Lipping, for taking me on as their student and guiding me through this study. They have been the ideal...