Page 171 - ITU-T Focus Group IMT-2020 Deliverables

ITU-T Focus Group IMT-2020 Deliverables 3

Master Slicing Controller: Is responsible for an entire country or large region. It talks to the individual C-RAN
and D-RAN slicing controllers and to the Metro & Core Network Controllers that govern the connectivity
between them.
C-RAN Instance Slicing Controller: Is responsible for a Cloud Radio Access Network. It talks to the Fronthaul
control, C-RAN Fabric Control, C-RAN server control and possibly to UE’s (indirectly). If PON is used for
fronthaul/backhaul then the slicing controller will talk to the PON controller. If PON is used to provide wireline
access services for home/enterprise then the C-RAN Slicing Controller may talk to the PON controller to
configure convergence between the fixed network and the wireless network by merging the relevant parts
of the packet core processing on both networks. When mobile edge computing is available, some of the
resources in the C-RAN can be assigned and interconnected for this purpose. There is no separate mobile
edge computing controller; all resources are considered usable for all functions (RAN/CORE/MEC) in the
CRAN. This does not preclude an ETSI MEC model, however it means an ETSI MEC model would need to be
virtualized so that its resources are not dedicated.
D-RAN Controller: Is responsible to talk to all the Distributed Radio Access Network devices. Essentially this
is a miniature C-RAN Instance Slicing Controller and is used when only a very few compute resources are
distributed close to the antennas. In this case the fronthaul is likely nailed up and non-configurable and the
fabric and processing controls are very limited; however some provision for mobile edge computing is
probably provided and configurable but as with CRAN there is no separate mobile edge computing controller;
the resources of the DRAN are usable for RAN/CORE and MEC as required. This does not preclude an ETSI
MEC model, however it means an ETSI MEC model would need to be virtualized so that its resources are not
dedicated.

Metro Network Controller/Core Network Controller: These controllers are pre-existing SDN or TSDN based
controllers which control the mesh of network elements that provide DWDM/OTN and Packet connectivity
on the scale of a metro or a country. Here we use the term “Core” in its networking not wireless sense.
FH slice control: This controller will ensure that a given Antenna is connected to the C-RAN fabric and
ultimately to the proper processing entities responsible for it. Since fronthaul can be analog, digitized analog,
mmWave, uWave, or even over packet, there are a lot of possibilities from complex, where there is a mesh
network to control, to trivial, where the interconnect is a point to point fibre. It is also possible that different
front haul technologies are used in different locations in the network based on fibre availability which may
imply different RAT split choices at different locations and possibly even at different times.
IDC slice control: This controller will manage the connectivity C/D-RAN to C/D-RAN and C/D-RAN to DC Pods.
This is most likely a transport SDN controller if the connectivity is OTN/DWDM, or an IP SDN controller if the
connectivity is IP/MPLS. It is possible there are two controllers if the connectivity is IP/MPLS over
OTN/DWDM. In the case where the IDC is implemented over a dark fibre then control is subsumed by the DC
or C-RAN controllers.

C-RAN Fabric Control: This controller manages the fabric inside the C-RAN. The C-RAN has very tight timing
requirements and therefore a normal DC Fabric Controller must be augmented. Clock distribution is required
as are control of pre-emption, scheduled Q’s and other very high performance packet behaviours. Possibly
bypass tunnel OTN/DWDM may be required within the C-RAN for delay/jitter requirements depending on
the architecture.

C-RAN Server Control: This controller manages all the processing entities in the C-RAN. These include general
purpose servers, servers with FPGA associated (hybrid), and likely non-standard acceleration hardware
required for special purpose digital signal processing. These servers may run VM’s, containers, bare metal
and may be used for RAT, MEC and CORE as required.

DC Controller: This controller is a pre-existing Data Center controller or hierarchy of controllers that are not
only dedicated to wireless. Their job of allocating connectivity between compute resources and of allocating
compute resources is simply re-used to allow control of 5G related packet processing entities to reside
outside of the C-RAN/D-RAN and to allow chaining of the relevant functions as dedicated by the packet core
chains for a given slice. The resources controlled in the DC are unlikely to be able to run RAT due to their
distance from the Antennas and the jitter/loss requirements; however they should be able to run MEC and
CORE equivalently with no distinction between the two.

165