Selective Enablement of L4S Transport for Latency-Sensitive Multimedia Delivery

The L4S standard has been gaining ground in the recent past, bringing together application and network service providers for building a more responsive internet. While application providers are working towards enabling L4S in their interactive applications, service providers are building dual queue architectures in their access networks. However, the state-of-the-art applies L4S to entire flows and may raise fairness considerations, esp. if there is deliberate abuse of the standard. While queue protection policing is currently also being developed, we propose novel methods for selectively enabling L4S, premised on judicious use of the low latency queue resource. In our proposal, we consider both ultra-low latency and low latency interactive video applications, wherein L4S is invoked based on PES picture size and CMAF chunk/fragment size, respectively. Further, we also contemplate transport layer handling with independent L4S and non-L4S QUIC streams and separate send/receive buffers when both pathways are used in conjunction for delivery. We conclude with recommendations of future work for further evaluation and development of our proposed methods.