Paper: D. Brunello, I. Johansson S, M. Ozger and C. Cavdar, "Low Latency Low Loss Scalable Throughput in 5G Networks," 2021 IEEE 93rd Vehicular Technology Conference (VTC2021-Spring), Helsinki, Finland, 2021, pp. 1-7, doi: 10.1109/VTC2021-Spring51267.2021.9448764.

intended to reduce queue delay problems, ensur-
ing low latency to Internet Protocol flows with a high throughput
performance. To reach this goal, it relies on Explicit Congestion
Notification (ECN), a mechanism that marks packets to signal
congestion in the network avoiding packets to be dropped

Low Latency Low Loss Scalable Throughput (L4S) is a
technology intended to mitigate the queue delay problem,
ensuring low latency to Internet Protocol (IP) flows without
affecting the throughput performance.

to change the Explicit Congestion Notification
(ECN) bits in the IP header as soon as the queues in the
network nodes start to grow.

it is possible to signal congestion early and to reduce the queue delay.

L4S provides an average
application layer throughput above the minimum requirements
of a high-rate latency-critical application, even at high system
loads.

In this paper, the implementation challenges of L4S in a 5G
network such as changing the ECN bit in the Radio Link
Control (RLC) layer and the coexistence of L4S with IP-
in-IP tunnels have been analyzed.

The latency can be adversely affected by events such as
congestion and packet loss.

To reduce the number of retransmissions and
to reach a higher throughput, the network nodes implement
relatively large buffers, where packets are queued instead of
being dropped

The research questions that will
be answered in this paper are the following:
• The first question concerns tunnels used in a cellular
network such as General Packet Radio Service (GPRS)
Tunnelling Protocol for user plane (GTP-U) and Internet
Protocol Security (IPSec) tunnels. Tunnels involve the
creation of new packets with new headers, and L4S uses
header bits to signal congestion. Do tunnels create any
obstruction for the L4S congestion signals propagation?
• Usually in New Radio (NR) packets are queued at the
RLC layer. Packets at the RLC layer are encrypted, thus
it is impossible to mark the packets to signal congestion.
Is there any solution to overcome this problem?
• How to handle classic and L4S traffic simultaneously,
is it necessary to use different schedulers with different
priorities?
• Given that L4S marks packets early when the queue starts

to grow (based on a threshold), what is the impact of
changing the marking thresholds?
• Can L4S improve key performance indicators (KPIs)
of high-rate latency-critical application traffic in a 5G
context?

AQM improves performance for end-to-end flows

If the traffic is
ECN-capable, AQM can mark packets instead of dropping
them.

TCP Prague [5], BBRv2 [6] and SCReAM [7]. These algorithms
adapt the rate of the sender node proportionally to the number
of congestion signals received.

It is decreased proportionally to the fraction of the CE-marked
packets received at the receiver node.

ECN is a standardized mechanism [2] for the management
and notification of congestion in the network, which avoids
dropping the packets in the buffer.

L4S [9] is a technology intended to ensure very low delay
for the IP traffic. It exploits ECN and is based on the
idea of marking packets as CE as soon as the queue delay
in the network node starts to increase above a threshold

if the packet
has codepoints ECT(0) or ECT(1), this means that the packet
supports ECN by the sender node, the network nodes can
change the ECN bits to CE to signal congestion.

thanks to the use of
scalable congestion control algorithms, it is possible to achieve
a good throughput performance and high link utilization even
if a high number of congestion events are signaled

The main
requirement for ECN to work correctly is that the network
nodes must be able to recognize the traffic that supports
ECN, and also be able to modify the ECN bits to signal the
congestion.

The steps of a communication session with the L4S support
are given as follow:
• The sender indicates L4S support through the use of
specific ECN codepoints (in this paper called L4S code-
points, which is explained in Table II).
• The network nodes recognize the packet as an L4S
packet. When congestion occurs, it is signaled by chang-
ing the ECN bits to indicate CE.
• The packet reaches the receiver, and if the ECN bits show
that congestion is experienced, the receiver notifies the
sender about the congestion.
• The sender, once notified about the congestion, reduces
the sending rate thanks to the scalable congestion control.

The receiver must be able to notice the
CE codepoint and notify the sender about the congestion

each node in the network
should be able to distinguish an L4S packet from a not-L4S
packet (also called classic packet).

L4S packets should have a reserved
queue where packets are CE marked as soon as the queue starts