Simulation Framework for HTTP-Based Adaptive Streaming Applications
Simulation Framework for HTTP-Based Adaptive Streaming Applications
复制标题
基于 HTTP 的自适应流应用程序的模拟框架
DOI：
--
复制
发表时间：
2017
期刊：
影响因子：
--
通讯作者：
A. Wolisz
中科院分区：
文献类型：
--
作者：
Harald Ott;K. Miller;A. Wolisz
基于互联网的视频服务的普及在过去几年中显著增加。基于互联网的视频点播的事实上的标准技术是基于HTTP的自适应流（HAS），其也越来越多地用于直播服务。HAS客户端的核心组件是自适应算法，它根据网络条件动态调整视频表示。与此同时，存在着大量的自适应算法的工作。不幸的是，许多实验研究缺乏彻底的性能评估。通常，原因是使用了不现实的网络环境，或与其他研究的结果不可比，或评估参数配置的子集太窄。我们认为，模拟的方法可以帮助解决这些问题，需要更少的努力来建立一个现实的网络环境，通过协助重现实验，并允许并行模拟，并可能运行速度比在真实的时间。目前的工作的贡献是设计和实现的一个基于HAS的应用程序，包括客户端和服务器端的仿真模型。它具有干净的模块化结构，允许轻松集成不同的自适应算法。客户端行为由一个可扩展状态机定义，该状态机可以轻松扩展以包含其他功能。此外，该模型提供了广泛的日志功能，用于监控体验质量（QoE）。我们将三种最先进的算法集成到模型中：FESTIVE，PANDA和TOBASCO 2。我们通过使用模拟室内Wi-Fi环境运行一组实验来证明该模型的有用性。
The popularity of Internet-based video services has significantly increased over the past years. The de facto standard technology for Internet-based Video on Demand is HTTP-Based Adaptive Streaming (HAS), which is also increasingly used for live services. A core component of a HAS client is the adaptation algorithm, which dynamically adjusts the video representation to the network conditions. Meanwhile, there exists a large body of work on adaptation algorithms. Unfortunately, many experimental studies lack a thorough performance evaluation. Often, the reason is the use of an unrealistic network environment, or incomparability of results with other studies, or a too narrow subset of evaluated parameter configurations. We argue that a simulative approach can help resolving these issues by requiring less efforts to set up a realistic network environment, by assisting to reproduce an experiment, and by allowing to parallelize simulations, and potentially run them faster than in real time. The contribution of the present work is a design and implementation of a simulation model for a HAS-based application, including both the client and a server side. It has a clean modularized structure allowing for an easy integration of different adaptation algorithms. The client behavior is defined by a Finite-State Machine that can easily be extended to include additional functionality. Moreover, the model provides extensive logging functionality for monitoring the Quality of Experience (QoE). We integrate three state-of-the-art algorithms into the model: FESTIVE, PANDA, and TOBASCO2. We demonstrate the usefulness of the model by running a set of experiments using a simulated indoor Wi-Fi environment.