Research On The Unfairness Problem In HTTP/2 Adaptive Streaming

HTTP Adaptive Streaming(HAS)over HTTP/2 has recently caught attention in the development of multimedia delivery.The server push feature of HTTP/2 enables the streaming server to send multiple video segments to the clients within a single request.Recent researches only investigate the performance of the HTTP/2 server push under a single-client scenario.In fact,when multiple clients share a limited bottleneck bandwidth,it is highly possible that the unfairness in bitrate selection will happen due to the mismatch of the segment download state among the clients.Meanwhile,the wrong estimation of the throughput by clients will result in the bitrate requests mismatch with the current available bandwidth,thus affects the Quality of Experience(QoE)seriously.In this thesis,firstly,the unfairness problem between multiple clients based on the HTTP/2 protocol server push mechanism is studied.Taking two clients as an example,this thesis analyzes various situation of ON-OFF Period mismatch between them.Experimental results show that the such kind of mismatch really causes the unfairness problem in practice;Therefore,a controller-based architecture is proposed to reasonably allocate bandwidth for each client and execute admission control.That guarantees the smooth playback of existing clients successfully;Finally,three representative throughput-based adaptive methods in HTTP/1.1 protocol are appliedto HTTP/2 server push mechanism.Experimental results show that all these three methods either overestimate or underestimate the throughput,leading to the wrong bitrate request,low bandwidth utilization,and rebuffing.A harmonic mean throughput adaptive optimization algorithm is proposed to solve the above problems.Experiments show that this optimization algorithm can measure throughput accurately,make full use of available bandwidth and reduce video quality fluctuation.