Video Traffic Modeling and Management over Video on Demand NetworksAbstract
With the popularity of multimedia applications over Internet, Video on Demand (VoD) is an important application because of its conveniences. For effective network content storage and delivery, Content Delivery Network (CDN) is always proposed such that video server farms are located at the metro networks.
Therefore, one of the major bottlenecks for transmitting video applications over Internet is at the border router of server farms. However, the border router always needs accurate traffic characteristics information provided by video sources for efficient bandwidth management. Up to now, many literatures have proposed some accurate video traffic description methods. However, these methods never consider the realistic situation that the peak data rate of access networks may probably be limited. In this thesis, we propose a new video traffic description scheme, Delay-Tolerant Multi-Leaky-Bucket (DTMLB). DTMLB is based on the deterministic traffic concept but does not use the worst-case model for analysis. DTMLB also introduces the consideration of different delay tolerance characteristics among I-frames, P-frames, and B-frames to improve the bandwidth utilization. Moreover, in our simulations, we also consider the realistic situation that the peak data rate is often limited by the available bandwidth bound of access networks. Simulation results show that when the peak data rate limitation is introduced, DTMLB can always provide lower transmission delay and higher bandwidth utilization than the other description methods. The description accuracy of DTMLB is also better than the other description methods in our simulations. |