Decentralized asynchronous optimization for dynamic adaptive multimedia streaming over information centric networking

By the envision of combing smooth viewing experience with high-efficiency content distribution, dynamic adaptive streaming (DAS) over information-centric networking (ICN) is becoming a promising trend for the future video services. However, optimizations of DAS flow transmission control and rate adaptation need to be revisited for better adopting the ICN with multicast, multi-rate forwarding and decentralized framework. In this paper, we propose a decentralized asynchronous method for ICN-DAS. We first formulate the problem as a two-stage optimization, wherein the first stage's objective is to optimize the transmission rate within network capacity constraints, and the second is adapting the video bitrate for the long-term viewing utility. A distributed asynchronous optimization algorithm (DAOA) is then proposed for solving the two-stage problem iteratively by a novel distributed switching mirror descent and virtual queue-based iterations. Analytic results including convergence, computation complexity and time-varying adaptation are provided to validate theoretically the DAOA's performance. Simulation-based testing has also been conducted for evaluating DAOA's performance and assess its viewing experience, in comparison with state-of-the-art solutions.