A scalable load generation framework for evaluation of video streaming workflows in the cloud

HTTP Adaptive Streaming (HAS) is increasingly deployed at large, gradually replacing traditional broadcast. However, testing large-scale deployments remains challenging, costly and error-prone. Especially, testing with realistic streaming loads from massive numbers of users is challenging and costly. To improve this, we introduce an open-source load testing tool that can be deployed in the cloud or on-premise in a distributed manner, for load generation. Our presented tool is an extension of an existing open-source web-application load-testing tool. In particular we have added functionality, that includes streaming load generation for a multitude of protocols (i.e. Dynamic Adaptive Streaming over HTTP (DASH) and HTTP-Live-Streaming (HLS)) and use-case implementations (e.g. live streaming, Video on Demand (VoD), bit-rate switching). The extension facilitates testing streaming back-ends at scale in a resource-efficient manner. We illustrate our tool's capabilities via a series of use-cases, designed to test, among others, how streaming deployments perform under different load scenarios, i.e. steep or gradual user ramp-up and stability testing over long periods.

HTTP自适应流（HAS）在大规模应用中日益普及，逐渐取代传统广播。然而，对大规模部署进行测试仍然具有挑战性、成本高昂且容易出错。特别是，使用来自大量用户的真实流负载进行测试既困难又昂贵。为了改善这种情况，我们引入了一种开源负载测试工具，它可以分布式地部署在云端或本地，用于生成负载。我们介绍的工具是对现有开源网络应用负载测试工具的扩展。特别是我们增加了功能，包括针对多种协议（即基于HTTP的动态自适应流（DASH）和HTTP直播流（HLS））的流负载生成以及用例实现（例如直播、视频点播（VoD）、比特率切换）。该扩展有助于以资源高效的方式大规模测试流后端。我们通过一系列用例展示了我们工具的能力，这些用例旨在测试流部署在不同负载场景下的性能，例如用户急剧或逐渐增加以及长时间的稳定性测试等。