Hierarchical video streaming for layered encoded video

Abstract

With the explosive growth of the Internet, the demand for various multimedia applications is rapidly increasing in recent years. Among different multimedia applications, streaming video is playing a very important role. In order to minimize the network congestion and server workload for video streaming, a video proxy located near client side is usually used to cache popular videos. On the other hand, as layered video can flexibly provide different quality of video according to the available bandwidth, it is more suitable for video streaming applications in a heterogeneous network environment like the Internet. This thesis presents the results of our work on enhancement and analysis of a hierarchical video streaming system using layered encoded videos. In this dissertation, we first investigate the problem of locking mechanism and derive a conditional locking process to solve the dilemma for proxy caching of layered videos. Two different proxy replacement algorithms, namely Least Popular Used (LPU) and simple count (SC), are then developed to cope with a dynamic changing environment for layered video streaming. The LPU algorithm exploits two different popularity parameters to obtain the most suitable choice on replacement processes while the SC algorithm takes the frequency of layers into consideration in doing replacements. Both of the algorithms can overcome the problem of cache pollution. Simulation evaluation shows that LPU and SC achieve a better cache performance when compared with other existing algorithms. Besides, a new intelligent mechanism is proposed for the video proxy caching algorithm to actively make decision when doing replacements in order to reduce the replacement frequency and proxy workload. Finally, a hierarchical architecture is explored for developing a cost-effective large scale on-demand video streaming system in a heterogeneous network environment using layered videos. An analytical model of the hierarchical streaming system is developed to optimize the cache efficiency of the proxy server under the constraints of network bandwidth in between the central and proxy server using the proposed SC caching technique. It is found that the blocking probabilities from the mathematical model closely match with the simulation results. If the proxy stores 10 videos and the arrival rate is 0.2 requests/s, about 75 Mbps network bandwidth is sufficient to serve customers that uniformly request different quality of videos with blocking probability less than 5%.