Congestion signaling system for adaptively marking packets in 5G radio access networks

Background

The rapid proliferation of interactive applications, such as virtual reality, cloud gaming, and videoconferencing, has shifted the performance requirements of modern telecommunications. These applications demand exceptionally low latency and consistently high throughput to function seamlessly over cellular last-mile connections, particularly within 5G Radio Access Networks (RANs). To meet these demands, network operators rely on sophisticated congestion signaling mechanisms like Low Latency, Low Loss, Scalable throughput (L4S). There is a critical need to extend these frameworks into cellular environments to ensure mobile users experience real-time responsiveness.

Despite their success in wired broadband, integrating these congestion mechanisms into cellular architectures presents significant challenges. Current approaches struggle with inherent protocol complexity and the strict opacity of internal RAN queueing, which hides critical congestion states from higher-layer Internet protocols. Consequently, standard active queue management systems fail to accurately estimate packet sojourn times amidst the rapid variability of wireless channels, leading to throughput underutilization. Furthermore, existing networks suffer from substantial feedback loop delays and round-trip time inflation caused by RAN jitter. Without timely queue state exposure, end-to-end algorithms cannot react quickly, resulting in detrimental bufferbloat and severe latency spikes.

Technology

This network architecture integrates Low Latency, Low Loss, Scalable throughput (L4S) congestion signaling into 5G Radio Access Networks (RANs). It introduces a processing layer within the Central Unit User Plane that passively estimates queuing delays using standard 3GPP F1-U feedback. By tracking packet timestamps, the system calculates queue egress rates and predicts packet sojourn times to apply adaptive Explicit Congestion Notification (ECN) marking based on traffic type. Furthermore, it minimizes feedback latency by short-circuiting the RAN, intercepting uplink ACK packets to update ECN fields before wireless transmission.

This solution is uniquely differentiated by delivering immediate congestion feedback without requiring modifications to underlying 3GPP or O-RAN protocols. Unlike traditional opaque cellular networks, it transparently exposes RAN queue states with negligible CPU and memory overhead. It significantly outperforms standard active queue management approaches in wireless environments, reducing one-way delay by up to 98% while maintaining near line-rate throughput. By dynamically adapting to channel variability and eliminating latency spikes, this lightweight, standards-compliant design seamlessly enables high-throughput, ultra-low-latency performance for demanding interactive applications over cellular connections.

Benefits

• Significantly reduces one-way network delay by up to 98% for interactive applications like AR/VR, cloud gaming, and videoconferencing.
• Maintains near line-rate throughput with negligible packet drops (under 1%), ensuring optimal network utilization even under heavy cellular congestion.
• Accelerates congestion feedback by “short-circuiting” the RAN, delivering immediate congestion state information to senders to prevent latency spikes.
• Integrates seamlessly using existing 3GPP and O-RAN control messages, requiring no modifications to underlying protocol structures.
• Operates with ultra-low system overhead, utilizing less than 2 microseconds of processing time per event with minimal CPU and memory impact.
• Improves the completion time of short-lived TCP flows by up to 4x without sacrificing the performance of long-lived flows.
• Ensures fair bandwidth distribution between L4S and classic traffic while remaining fully compatible with advanced 5G features like MIMO, network slicing, and carrier aggregation.

Applications

• 5G cloud gaming
• Mobile AR/VR streaming
• Low-latency mobile videoconferencing
• Real-time remote teleoperation
• Industrial IoT real-time control

Additional info

This system integrates L4S signaling into 5G RANs via a Central Unit User Plane layer. It passively estimates RLC queuing delays using F1-U feedback and a packet profile table to predict sojourn times. By applying adaptive ECN marking and short-circuiting feedback through uplink ACK modification, it delivers immediate congestion data to senders. This architecture optimizes interactive application performance while maintaining compatibility with existing 3GPP protocol structures.