Answers are grounded in this paper and cite their sources.
The growing importance of video conferencing and the demand for improved quality of experience (QoE) highlight the need to understand the underlying architectures of multi-party conferencing systems. This work surveys real-world conferencing architectures and classifies them into seven types: Relay, Selective Forwarding Unit (SFU), Multi-point Control Unit (MCU), Peer-to-Peer (P2P) Relay, P2P SFU, P2P MCU, and P2P Mesh. Additionally, we present uvgComm, a video conferencing platform supporting multiple conferencing architectures, provide a systematic methodology for evaluating conferencing architectures, and propose a new hybrid architecture based on the findings of this work. In uvgComm, SFU and P2P Mesh architectures are combined to implement the proposed hybrid architecture, which reduces client-to-client (C2C) latency through latency-aware path selection while respecting upload-bandwidth constraints. The SFU, P2P Mesh and hybrid architectures were evaluated in a virtualized environment using an AMD Ryzen Threadripper PRO 7995WX processor with 2 to 16 clients, a 4K base resolution, simulated real-world latencies, and a 6.0 Mbit/s per-client upload-bandwidth limit. Our results show that the proposed hybrid architecture can utilize additional upload bandwidth to reduce latency by 30–41 ms while maintaining equivalent video quality to that of SFU. For 16 clients, this corresponds to approximately 88 % of the latency gap between SFU and P2P Mesh. These findings demonstrate the potential of hybrid path selection to improve QoE under varying network conditions.
No takes yet. Share an insight, caveat, or question.
Räsänen et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: