Research On Proxy Caching Algorithms For Streaming Media

With more and more multimedia applicantion, many researches have attached importance to the stream media that can provide the capacity of"Downloading and Playing"and allows people to enjoy the continuous media on line. With popularizing the typical stream system, such as Video-on-Demand, Video Conference and E-Learning etc, the multimedia traffic takes the largest portion in the current Internet traffic. However, there is some obvious difference between stream media and non-stream media. a) Stream media will consume more bandwidth than non-stream media; b) Clients will be more sensitive on startup delay when they use stream media; c) Heterogeneous clients will ask different quality requirement for received media according to their own environment. This brings new challenges and pressures to networks and stream media technology in the following fields, including the assurance of quality of service for stream media, the assurance of stream media system performance and the influence of heterogeneous client behavior and network environment on stream media transmission.To solve the above problems through providing network edge service capacity, this thesis addresses on the proxy cache of stream media. The content of this thesis includes four issues: (1) modeling and analysis of stream media internal popularity distribution; (2) prefix stream media caching algorithm based on internal popularity distribution model; (3) prefix length and quality two-dimension adaptive stream media caching algorithm. The works in this thesis has been supported by the Natural Science Foundation of China"Investigation of Wireless Multimedia Techniques based on Multimedia Transmission Property"(No.60202005) and"Internet-based interactive streaming delivery technique"(No.60302004) respectively, and also supported by the Australian Research Council Foundation"Interactive video-on-demand for e-Learning"(No.LX0240468) and Natural Science Foundation of Hubei"Stream media caching based on quality and time two-dimension attribute"(No.2005ABA264). The contributions of this thesis include:1) In order to find the rule of content popularity distribution, a k-transformed Zipf-like model has been proposed through analyzing the real trace from three different servers in two different web sites in this thesis. The analysis verifies the correctness of the model. And we further analyize the effect of model parameters.2) In order to have a good tradeoff between the utility of cache space and the accuracy of internal popularity distribution, a new segmentation caching algorithm which is based on the k-transformed Zipf-like model has been proposed. Through evaluating and updating key parameter of model in an efficient way, this algorithm can get the accurate internal popularity distribution even using little cache space to record the access information and get a good cache performance.3) With the limitation of cache space, improve the stream media transmission quality and lower the consumption of backbone bandwidth is a conflict. A new stream media caching and transmission scheme which combines with FGS coding has been proposed in this thesis. The results of simulation show that this scheme can let the clients get the best quality of stream media when satisfy the requirement of backbone bandwidth consumption.