SVDN: packetization and layer synchronization for scalable video delivery over peer-to-peer networks

Abstract

Layered video streaming can achieve dynamic video quality adaptation according to end user device capabilities, network conditions, and user preferences. In layered video streaming, synchronization between different video layers and packetization are two essential design challenges to facilitate video quality adaptation. This paper proposes SVDN--a novel Real-Time Scalable Video Delivery system using Peer-to-Peer (P2P) Networks--and focuses on the following perspectives: (1) a distributed video layer synchronization algorithm for P2P networks is proposed, which is the key algorithm facilitating dynamic video quality adaptations for live video streaming; (2) the corresponding packetization design for fixed chunk size is presented, which can fully utilize chunk space. Both Variable Bit Rate and Constant Bit Rate are supported by the packetization method; (3) an adaptive chunk size determination algorithm is proposed to estimate a fixed chunk size for each video layer, which further improves the packetization performance. Evaluation of the implementation shows that SVDN significantly improves the bandwidth usage when compared with existing packetization solutions for fixed chunk sizes. The average minimum packetization overhead of the state of the art solution ranges from 8.5 to 15 %, while SVDN can achieve virtually zero padding overhead and at the same time fully support spatial, temporal and quality layer switching of H.264/Scalable Video Coding encoded video.