Page 302 - 5G Basics - Core Network Aspects

1 Core network aspects

transmission flows of a service that share a bottleneck link are required to be coupled for congestion control.
In terms of the detection mechanism, the sender may inject probe packets as active detection techniques or
employ application and/or transport data and acknowledgement (ACK) packets for measurement purposes
as passive detection techniques. In addition, loss-based techniques rely on loss characteristics within and
across sub-transmission flows as signals of congestion, and similarly, delay-based techniques use delay
characteristics and changes within and across sub-transmission flows. The sub-transmission flows that share
a common or separate bottleneck are required to be considered regarding which of the packets from these
two flows are likely to be dropped or delayed in close time.

In contrast to single-path transmission, services can be separated and transmitted simultaneously so that the
sub-transmission flows of a service transmitted through distinct paths are required to be coordinated. With
the aid of multi-path transmission control, some of the traffic can be shifted from more congested paths to
less congested ones, thus compensating for lost bandwidth on some paths by moderately increasing
transmission rates on others. The congestion control mechanism of multi-path transmission is required to
couple the paths in order to shift traffic from more congested to less congested paths, and thus standard TCP
congestion control mechanisms are not applicable.
Therefore, alternate congestion control mechanisms are necessary, such as, fully coupled congestion control,
coupled congestion control or semi-coupled congestion control, which take the sub-transmission flows of a
service into consideration. In case of path loss, path failure detection is necessary, and the traffic on the failed
path should be scheduled and transferred according to the available paths.

7.3.3.2 Throughput guarantee mechanism
In multi-path transmission, if the end-to-end delay and bottleneck bandwidth are not properly addressed,
there may be many packets along multiple paths, which can arrive late and can lead to a large number of
out-of-order packets at the receiver, and can eventually cause serious degradation to QoS.
In order to solve the above-mentioned problem, a buffer is required to be used to execute packet reordering
at the receiver. At the same time, packet delivery order on each path is required to take both bandwidth
aggregation and end-to-end delay into account in order to obtain the optimal transmission throughput.
Several mechanisms can be used according to the available bandwidth and end-to-end delay on each path,
such as the earliest delivery path first (EDPF) mechanism.
Duplicate ACKs or retransmission timeout (RTO) are two methods for triggering retransmission. The
retransmission timer is adjusted dynamically, based on the measured RTTs. In multi-path transmission,
packet reordering or other events that bring a sudden change of the RTT may lead to unnecessary
retransmissions, which would cause the sender to wait a longer time for an acknowledgment. In order to
avoid unnecessary retransmission, RTO of a path is required to be calculated appropriately; refer to [b-IETF
RFC 5682].
In standard TCP, every lost packet is recovered through retransmissions at the transport layer without coding
techniques. In multi-path transmission, reliability can be achieved by introducing redundancy in the form of
copies of each packet sent through multiple paths. However, network coding can recover original packets
from receiving packets even if there is partial information loss in the packets. In addition, the application of
network coding can improve network throughput, reduce energy consumption, reduce transmission delay,
and the encoded data is secured in transmission. In general, network coding schemes can be divided into
linear coding and nonlinear coding.

7.3.3.3 Efficient retransmission mechanism
In standard TCP, a single-path algorithm is adopted with traffic retransmission for packet loss. However, to
ensure the delivery of packets on time with acceptable quality, there are several retransmission paths
available other than the original one in multi-path transmission environment. In some cases, retransmission
through the original path is inefficient, especially when the original path is in a state of high load. In multi-
path transmission control, retransmission through the original path is not a strict constraint, and choosing
the retransmission path mainly depends on the QoS requirements of the retransmission traffic.

292