Joint Routing and Layer Selecting for Scalable Video Transmission in SDN

Scalable Video Coding (SVC) is a promising technique to enable flexible video transmission for a time-varying available bandwidth. Most of previous works about the scalable video transmission independently focus on the optimal number of video layers to be transmitted or the routing of each video layer. An emerging network technology, Software Defined Networking (SDN), can utilize application-information for network programming so that their combination is available. Therefore, this paper proposes a joint algorithm to concurrently decide the number of video layers to be transmitted and the transmission path of each selected video layer in SDN. The joint decision problem is formulated as a Markov Decision Process (MDP), which takes into account the time-varying non-SVC traffics in each link among all candidate paths from the sender to the receiver. The optimization objective is to achieve a good tradeoff between the visual quality of the transmitted SVC videos and the packet loss of non-SVC traffics. Finally, Q-Learning algorithm, which does not require any priori information of the transition matrix, is employed to solve the MDP optimization problem and derives a dynamic joint control strategy. To verify the performance of the proposed method, we conduct our experiments in Mininet. The results demonstrate that our algorithm performs better in the long-term reward than the baseline algorithms.