Enhancements to 5G Data Transfer – Part 1

As 5G continues to mature and the first wave of 5G Advanced deployments emerge, the focus of innovation is shifting from raw speed and headline latency figures to how the network behaves under real-world conditions. The next generation of connectivity is about more than peak throughput; it’s about delivering consistent, predictable performance across a growing range of demanding applications.

From immersive XR (eXtended Reality) and industrial automation to cloud gaming and ultra-reliable control systems, today’s services place unprecedented demands on mobile networks. Meeting these expectations requires not only improvements in the radio layer but also significant evolution in how data is handled within the network itself.

This two-part blog explores how enhancements to data transfer are shaping the future of 5G. In Part 1, we focus on developments within the 5G Core and RAN, specifically how innovations such as L4S, user plane redundancy, multi-access PDU sessions and ATSSS, and network slicing are improving the way data flows through the network. In Part 2 we will look at the supporting ecosystem, including transport network evolution and Multi-Access Edge Computing (MEC).

L4S: Low Latency, Low Loss, Scalable Throughput

Today’s immersive multimedia experiences place significant demands on networks, relying on low latency to deliver responsive, real-time interactions. Traditional methods of managing congestion in networks tend to focus more on throughput rather than latency and are poorly optimised for use with delay-critical traffic.

When traffic for these applications encounters congestion in the network it can lead to long queuing times as delay-critical packets get stuck behind standard traffic, this can lead to packet loss and an unsatisfactory quality of experience for the end user. Whilst existing QoS solutions can prioritise low-latency traffic, bottlenecks in the network can remain.

L4S (Low Latency, Low Loss, Scalable Throughput) introduces a new approach. Instead of waiting for packet loss as a sign of congestion, L4S uses ECN (Explicit Congestion Notification) to signal potential congestion earlier. Combined with dual-queue coupled Active Queue Management (AQM), L4S allows delay-sensitive traffic to bypass long queues while still maintaining high throughput for other traffic types.

Figure 1 - Key Use Cases for L4S

L4S in the 5G Network

Support for L4S was introduced in 3GPP Release 17, aligning with IETF standards (RFC 9330–9332). ECN marking can now occur at several points in the network, including the NG-RAN, UPF, and non-3GPP access. This proactive approach to congestion marking  is particularly valuable for real-time services such as XR, cloud gaming and low-latency financial data, which demand consistent performance even under fluctuating network conditions.

• Uplink traffic (e.g. video conferencing, live telemetry, live streaming) benefits from ECN marking to prevent queue build-up at the gNB or UPF.
• Downlink traffic (e.g. cloud gaming, real-time data feeds) experiences reduced jitter and smoother delivery.
• Bidirectional traffic (e.g. collaborative apps, automation systems) gains end-to-end latency consistency when ECN is applied in both directions.

ECN Marking in the 5G Core

The implementation of L4S in 5G allows for ECN marking to take place either directly on the RAN, within the core at the UPF, or in the case of Non-3GPP access at the N3IWF/TNGF.

Typically, ECN marking is performed at the UPF, as it is well positioned in the traffic path to inspect IP headers. However, since the bottleneck usually occurs on the RAN side, the RAN can provide congestion information to the UPF to support more effective ECN marking. This centralised approach ensures consistent behaviour across the network and maintains signalling even if marking is not applied at the RAN itself.

Figure 2 - L4S in 5G Overview

Establishing an L4S QoS Flow

L4S operation is enabled at the QoS Flow level. Although no dedicated 5QI values are defined for L4S, low-latency 5QIs are typically used. Detection relies on ECN field values such as ECT(1) which signals support for ECN marking.

Once an application starts sending ECN-marked packets, the network can detect this traffic and establish a dedicated QoS Flow optimised for L4S. The marking is then applied dynamically based on local queue conditions or congestion reports, allowing the system to continuously adapt.

Strengthening Resilience: User Plane Redundancy

While speed and latency dominate much of the discussion around 5G, reliability and availability are just as important. Many critical 5G use cases – including industrial control, remote surgery and connected vehicles – cannot tolerate packet loss or service interruptions. This is where user plane redundancy becomes essential.

Originally introduced to support URLLC (Ultra-Reliable Low Latency Communications), user plane redundancy ensures that data can continue to flow even if part of the network fails. Operators can choose where redundancy is applied:

• End-to-end redundancy: The device sets up two separate PDU sessions over independent paths, connecting through different gNBs and UPFs but ultimately reaching the same data network.
• N3 redundancy: Even with a single base station, redundant paths are established for the N3 interface, helping maintain service continuity if the backhaul is unreliable.

Figure 3 - User Plane Redundancy

While this duplication increases data load – potentially doubling traffic for some devices – it is often essential for services that cannot risk downtime. As demand for ultra-reliable connectivity grows, redundancy within the core and RAN becomes a critical design consideration.

Multi-Access PDU Sessions and ATSSS

The rise of multi-access devices capable of connecting over both 5G and Wi-Fi introduces new possibilities for performance and resilience. Multi-Access (MA) PDU sessions allow a single session to use both 3GPP and non-3GPP access simultaneously, enhancing throughput and reliability.

At the heart of this capability is ATSSS (Access Traffic Steering, Switching and Splitting), which determines how traffic is distributed between networks:

• Higher layer steering uses protocols such as MPTCP or MP-QUIC to split traffic above the IP layer.
• Lower layer steering distributes traffic based on performance metrics such as round-trip time or packet loss.

Figure 4 - ATSSS-N4 Rules

Once an MA-PDU session is established, traffic policies – defined by the network and delivered as ATSSS Rules – govern how data is handled. Modes include:

• Active/Standby: One access network is used unless it becomes unavailable.
• Smallest Delay: Traffic is directed to the network with the lowest latency.
• Load Balancing: Traffic is split across networks based on current conditions.
• Priority: Traffic is sent over the preferred network unless congestion occurs.

ATSSS improves both performance and reliability, ensuring users experience seamless service even as network conditions change.

Network Slicing: Customising Connectivity

Not all 5G services have the same requirements. Some demand ultra-low latency, while others prioritise bandwidth or security. Network slicing allows operators to divide a single physical network into multiple virtual networks, each optimised for specific use cases.

Once a device is authorised for a slice, a PDU session is established across that slice, enabling tailored data treatment such as enhanced security or prioritised scheduling. This granular control allows operators to support diverse services – from consumer applications to mission-critical industrial operations – on the same infrastructure.

Looking Ahead

The enhancements explored here – L4S, redundancy, multi-access sessions and network slicing represent fundamental changes in how data is managed within the 5G Core and RAN. Together, they enable networks to deliver consistently low latency, high reliability and flexible service differentiation.

In Part 2 of this series (arriving 3rd November), we will shift focus beyond the core to explore how transport network evolution and Multi-Access Edge Computing (MEC) are extending these capabilities and enabling a new generation of applications at the edge of the network.

Would you like to learn more about L4S and how it is used in 5G?