摘要近年来，基于 HTTP 的动态自适应流媒体(DASH)已经被广泛部署在因特网上。然 而，由于延迟问题，DASH 在直播领域的应用仍十分有限。通过利用全双工兼容超文 本协议中的服务器推送机制，通过减少片段时长将低延迟，通过利用 K-PUSH，即一 个请求对应多个片段，维持可控数量的请求报文避免报文膨胀，FDH-DASH 为解决延 迟问题提供了可能。然而，由于直播流媒体的低延迟限制，以及因此带来的小缓存 特性，为 FDH-DASH 设计平滑的码率切换策略仍是一个极具挑战的问题。本文为低延 迟直播流媒体场景设计了一种基于动态 K-PUSH 的码率自适应算法。通过动态的决定 每个推送会话的 K 参数，可在带宽变化反应速度和减少请求报文方面取得较好的平 衡。本文将所述码率自适应问题归结为一个多目标优化问题，并构建了一个多目标 优化模型，包含一个基于双阈值的约束条件机制以避免缓存的上溢和下溢，和一组 精心设计的优化目标函数以最大化用户体验质量。最后，我们给出了所述多目标优 化问题的求解算法，包含一个基于穷搜机制的优化算法以求解出所有的 Pareto 最优 解，和一个基于动态线性化方法的决策算法以为每次推送会话选择最终的 K 和 R。理 论分析以及实验环境和真实网络环境的扩展实验均证明了所述方法的高效性。 73792

毕业论文关键词 动态 HTTP 流媒体(DASH) 直播流媒体 码率自适应  全双工兼容超文本协 议(FDH)  K-PUSH

Title The study of dynamic K-push based rate adaptation  for HTTP/2 live adaptive streaming           

Abstract Abstract—Recently, Dynamic Adaptive Streaming over HTTP (DASH) has been widely deployed over the Internet。 However, due to the latency problem, the application of DASH in the live area is still very limited。 By leveraging server push technology in Full Duplex HTTP-compatible protocols (FDH), FDH-DASH provides an opportunity to address the live latency problem by reducing segment duration while still maintaining a controllable number of requests using K-push, i。e。 one request every k segments。 It is, however, a challenging problem to smooth video bit-rate switching for FDH-DASH due to the low latency constraint and thus small buffer size in live streaming。 In this paper, we propose a dynamic K-push based rate adaptation scheme for FDH-DASH。 By dynamically determining the number of segments for each push session, a good balance between reaction speed and request overhead reduction is achieved。 We formulate the rate adaptation approach into a multi-objective optimization problem。 We build the multi-optimization model combining a dual-threshold based constraint mechanism to ensure no buffer overflow/underflow during each push session with the carefully designed objective functions to maximum the quality of user experience。 A novel algorithm is designed to solve above multi-optimization problem, which includes an exhaustive search method to solve all pareto optimal solutions and a decision-making scheme to find the best K and video bit-rate for each push session。 Theoretical analysis and extensive experiments on our network testbed and the Internet demonstrate the good efficiency of the proposed method。 

Keywords Dynamic adaptive streaming over HTTP (DASH), live streaming, rate adaptation, full duplex HTTPcompatible protocols (FDH), k-push 

目 次

1 绪论 1