基于多个网络接口的DASH系统设计与实现

摘要

对基于HTTP协议的动态自适应流媒体来说,客户端可以根据不同的网络状况自适应地调整请求的视频比特率.当前DASH的研究主要集中于单链路的比特率切换算法的设计.而考虑到移动终端拥有多个网络接口以及短视频的播放流行度,开发一个基于LTE/Wi-Fi的多接口DASH系统——PASM(Predictive Adaptive Streaming system over LTE/Wi-Fi Multiple access networks).PASM从两个互补方面提高了DASH性能,一方面,减少了由于两条链路异构性造成的视频块的乱序到达,充分利用两条链路的带宽之和进行DASH传输,提高了请求视频的质量;另一方面,在剩余请求视频时长不多时,采用一种更加激进的视频比特率请求算法,减少了带宽的不充分利用,提高了视频的平均比特率.实验结果表明,在各种链路带宽差异的网络下,PASM保证了视频比特率的平滑切换,同时提高了视频平均比特率.%In dynamic adaptive streaming over HTTP, a client is able to switch between several video bitrates to cope with varying network conditions.The state-of-the-art adaptation algorithms are mainly designed to serve DASH flow on a single link.In view of multiple wireless interfaces equipped in mobile devices and the dominance of short Internet videos, we develop a predictive adaptive streaming system over LTE/Wi-Fi multiple access networks, namely PASM.PASM improves DASH performance through two orthogonal but complementary approaches.On one hand, PASM aggregates the bandwidth of LTE and Wi-Fi links for better video qualities, while inducing out-of-order chunk arrivals due to varying and heterogeneous bandwidth of each link.On the other hand, when the remaining video length is relatively short, we adopt a more aggressive algorithm to determine which chunk bitrate should be chosen, which leads to the increased average bitrate and reduced the number of bit-rate switching.Experimental studies show that PASM can guarantee the smooth video bitrate switching, and meanwhile increases the average video bitrate in various network conditions.