Fine-grained Transmission Optimization of Large-scale Web VR Scenes

Abstract

The latency of transmitting large-scale WebVR scenes over mobile Internet is known as the bottleneck problem. This paper tries to challenge this problem by combining adaptive packaging transmission, use UDP and QUIC protocol for transmission instead of TCP and HTTP2. In addition, we also propose to use p2p for transmission to reduce the load pressure of the server during data transmission. Different with those pure research DVE (Distributed Virtual Environment) P2P works built on simulation platform, a novel WebVR-P2P framework is realized based on WebTorrent and WebGL. On server side, large- scale WebVR scenes are divided into smaller fine-grained subspaces in terms of closeness and visibility to lower networking congestions. These two preprocessing steps are integrated to decrease less bandwidth occupation at utmost. Then, each fine-grained subspace is packaged adaptively in terms of Frustum Fill Ratio (FFR) for smooth and efficient transmission. A new WebTorrent framework is extended to transmit Web3D files and all packaged WebVR subspaces are transferred in the peer-to-peer style. At the same time, on server side, we introduced the support of QUIC, and improved the support of QUIC in the project. Finally, WebVR-P2P and QUIC supported platform is implemented based on all above key technologies, a large-scale WebVR scene (an industrial park: 1325M) is chosen to test for P2P transmission and QUIC transmission performance in this WebVR-P2P platform, the practical experimenting results are conducted to show the effectiveness and potentiality of our proposed solution.