To QoE or not to QoE
Szilveszter Nádas, Lars Ernström, László Szilágyi, Gyanesh Patra, Dmitri Krylov, Jonathan Lynam · 2024
Conventionally, intermediate network nodes and congestion control protocols cooperate to allocate bitrate roughly equally between traffic streams (all else being equal). Respecting the layering principle, a network had no basis by which to determine the incremental value of allocating more bandwidth to one stream vs. another. Furthermore, making these sorts of decisions was considered unnecessary during epochs where adding capacity was easier than adding complexity. In this paper, we reconsider these perspectives due to the rise of real-time media streaming as a key enabler for many advanced 5G applications, and the extra pressure it puts on network capacity. Although real-time streaming presents many challenges, e.g. high bitrates and tight latency requirements, it also tantalizingly offers an approach to determine the relative value of bandwidth allocations over short timescales, through dynamic monitoring and management of Quality of Experience (QoE) of each stream. In this paper we show that QoE-aware dynamic network resource allocation is possible and can produce strong gains in the numbers of simultaneous streams and stabilize user QoE. We propose further research in this area both on how to create the most efficient mechanisms to realize the demonstrated gains as well as on how to integrate these mechanisms into existing network and software architectures.