Dynamic Adaptive Streaming over HTTP (DASH) technology has recently gained popularity as client-driven streaming solution based on HTTP. It addresses the increasing and heterogeneous demand of multimedia content over the Internet by offering several encoded versions for each video sequence. Each version is encoded at a different bitrate and resolution so that each user can select the most suitable version depending on the video client capabilities and network bandwidth. Thus, the DASH controller, responsible for the selection of the representations to download, entirely resides at the client side, leading to a distributed system. The main advantages of this distributed technology are its extreme scalability, its ability to work with already deployed web server technologies, and its ability in accommodating for clients heterogeneity.
Taking a provider’s perspective, we mainly investigate the role of optimal representations set stored at the server. The only existing guidelines for selecting the parameters of the video representations are recommendations from system manufacturers. However, they are not necessarily adapted to the experienced network and users population. Our research focuses on optimizing the set of representations that should be generated by the ingest server and show that the existing recommended sets have critical weaknesses. Relavant publications can be found here.
At the client side, we are also interest in improving the behavior of the client controller. The main weakness of existing controllers are (i) the quality fairness among users is not necessarily achieved by existing clients models; (ii) the optimal behavior should be adapted real-time to the expericenced context, characterized by both networks and content information. With our research, we target to face both challenged studing learning strategies for clients’ behavior aimed at reaching quality fairness among users.