High-Performance Virtual Reality Services via Multi-network Interface

Recent advancements in hardware technology and high-speed Gigabit WiFi solutions have paved the way for sig-nificant enhancements in Internet throughput, thereby enabling more sophisticated applications such as the combination of live streaming and virtual reality (VR) and augmented reality (AR) applications. Despite the growth, VR/ AR live broadcasting has a notably high computational demand for video stitching and encoding. Traditional methods rely on local resources, which are often insufficient. The advent of cloud computing offers a promising solution by outsourcing compute-intensive tasks to the cloud, thus alleviating the load on local systems. When uploading content to the cloud, a primary issue to address is how to ensure that footage captured from different cameras can arrive as synchronously as possible to effectively reduce processing latency. In this paper, we propose a novel mechanism that employs a multi-network interface controller at the acquisition terminals to enhance bandwidth utilization and ensure the synchronous arrival of multi-camera feeds for effective stitching. A multi-path transmission model is designed to manage potential network congestion and optimize buffer queue lengths in the cloud. Besides, a dynamic routing strategy is proposed to optimize the data forwarding by observing the instant network status. Our approach is validated through extensive simulations and prototype system experiments, demonstrating superior resource efficiency and performance over existing methods.