Construção de um serviço conteinerizado de vídeo sob demanda baseado em DASH para experimentação e coleta de métricas de desempenho

Abstract

The aim of this research is to study an on-demand video distribution environment. As the deployed service goes through distinct conditions during the time it is running, a platform component gathers data related to its overall performance. The collected data set will then be made suitable for future investigations where machine learning will be used. The methodology is based on carrying out controlled experiments using a video streaming service with support for dynamic adaptation by applying the DASH method, deployed in a conteinerized environment, accessed by a video client that is able to register a set of pre-defined metrics. As the video client captures the aforementioned metrics, a load generator tries to affect, in a controlled manner, the operation of the plaftorm. We show that the built platform was able to appropriately stream every DASH configured video available, while the load generator made the dynamic change in video quality possible during the experiment, resulting in a broader diversification of the data held within each newly obtained data set.