Research On DASH Scheduling Algorithm Based On HTTP/2 And SVC Coding

With the popularization of Internet and diversification of user needs,many multimedia services such as video conferencing,online education,and live broadcast spring up like mushrooms.In our country,live video broadcasting accounts for 27%of the time spent by mobile Internet users,which means video traffic becomes an important part of web traffic affects every aspect of human live.Media streaming technology is an important means to carry video traffic in the Internet.As an important streaming framework,DASH has gradually been accepted by major Internet service providers.In this thesis,we improve the existing DASH-based solution from two aspects:(1)Adaptive Bitrate Algorithms(ABR);(2)Transport Protocol.The first improvement is the adaptive bitrate scheduling algorithm.In the DASH architecture,due to the high bitrate fluctuation of VBR encoding,the average bitrate in the MPD file often cannot accurately represent the size of a specific video chunk.Scheduling algorithms which make decisions based on the average bitrate will affect user QoE due to chunk size overestimation or underestimation.Based on SVC encoding,this thesis designed a size-prediction algorithm based on the size correlation between different layers of video,and enhanced traditional adaptive algorithms with a size predictor.Results showed that compared with using the average bitrate in MPD file,the size predictor proposed in this thesis was closer to the actual size of the video chunk.The average playback bitrate increased by 10%?20%.Other QoE metrics had also improved.The second improvement made is for protocol.DASH over HTTP/1.1 has many inevitable drawbacks,especially when SVC encoding is used.In order to obtain a playable video segment,multiple HTTP requests are sent for each layer to enhance video quality,which wastes multiple RTT time.When network latency suddenly becomes higher,the additional latency will cause buffer starvation and playback stall.Standardization of HTTP/2 provides an opportunity for us to alleviate the aforementioned problem.In this thesis,we first found that some new features of HTTP/2 can be used to improve video streaming quality.We then augmented the client-side scheduling with HTTP/2 server push and combined it with four traditional algorithms.Results showed that augmenting the scheduling algorithm with HTTP/2server push can significantly improve the average playback bitrate and reduce start-up delay.Our results demonstrated that server push can be added to traditional ABR algorithms to increase user QoE.