QCON: Seamless QoE-Aware 5G Streaming via Multi-Connectivity

Mobile real-time video streaming (RTS) applications-cloud gaming and AR/VR-require consistent high throughput and low latency to satisfy user Quality of Experience (QoE), yet today's wireless links fluctuate wildly. While multi-path solutions seem promising to tackle such single-link fluctuations, existing transport-level solutions require multiple cellular subscriptions, which most users don't have. In this paper, we leverage 5G multi-connectivity, which allows simultaneous connection to multiple base stations (e.g., 5G and 4G) and is already deployed in commercial networks. However, our measurements show RTS applications still suffer from single-link fluctuations due to operators' deliberate policies restricting multi-connectivity to conserve 4G backup links regardless of application demands. To optimize application QoE while respecting operator policies, we present QCON, a QoE-driven multi-connectivity solution that efficiently utilizes backup links based on precise application QoE. For practical deployment, we design QoE Monitor to infer application QoE within the RAN and develop multi-link scheduling to optimize both QoE and radio resource efficiency. We also design prioritybased re-injection utilizing RAN link recovery mechanism to prevent video stalls. Our prototype implementation of QCON on a RAN intelligent controller within an Open-RAN testbed demonstrates 2.1× improvements of bitrates, enhancing tail frame rates by 4-5× with efficient backup link use compared to existing multi-link scheduling schemes.