Page 175 - ITU-T Focus Group IMT-2020 Deliverables

ITU-T Focus Group IMT-2020 Deliverables 3

For example we may want to look at the utilization in a particular set of sectors in the network and if they
are above some threshold and growing we may want to look for another sector in a lower priority slice where
we can steal some resources, reassign them to the higher priority slice in the sectors in question allocate new
CORE CPU, interconnect it to the RATs and then when things calm down, return the resources. In order to do
this the orchestration system will need to allow the user to specify the appropriate conditions including which
sectors to check, how to determine utilization, look at the trend in utilization, check it against a threshold,
then look for low utilization sectors in lower priority slices, then determine which has more resources than
necessary (or can be sacrificed), then compute a delta to the current higher priority slice and finally apply the
resources.
We can imagine a long list of things that can be queried, of Boolean operations and of a primitive set of
actions must be defined in order to implement such a tool kit. Various interfaces for Java/Python/C++ etc.
would need to be provided.

In addition to match/action type behaviours there is also a need for various auditing functions. It would seem
likely that the orchestration system should audit itself and the network for correct functioning via various
assertions.

ASSERT: slice(“slice-A”).sector(*).frequencyRange().notOverlaps(2.60Ghz, 2.90Ghz) ;
It’s very likely that two independent orchestration systems will be required. One which can perform the
actions and another which simply audits and ensures that high level rules of behaviour are in fact being
honoured and issuing warnings or taking punitive actions if they are not. One example would be to monitor
spectrum use to ensure that no license rules are being violated. Since the various RATs are highly
programmable it would be relatively easy to accidentally transmit on the wrong band in the wrong location
or wrong power level. An auditing system would allow for the specification of rules which must at all time be
met and make it less likely the system would operate for too long in an unsafe or non-compliant manner.
Auditing could be done either at the time of specification or at the time the change is actually implemented,
or post change.
8.2.5 Cookies

“Cookies” is a term used to describe data which is stored by a second software entity on behalf of a first
software entity without really knowing what it is. Cookies are useful because they allow state to be
maintained even when the first software entity goes away completely.

The various components that the orchestration system talks to should store cookies on behalf of the
controller sub-systems associated with the various objects under their control. This will allow an
orchestration system to resynchronize with manual or previously automated slices and greatly reduce the
burden of migration and auditing. For example, resources associated with a particular slice could be
associated with a cookie whose key is the name of the slice, this includes all the end to end resources such
as antennas, fronthaul, C-RAN fabric, C-RAN compute, etc. and all could carry such tags. This is also a very
useful debugging tool and would allow lower level query as to what resources were being used for without
talking directly to the higher layer orchestrator.

8.2.6 Control plane enhancement for 5G transport network

8.2.6.1 Control plane for low latency, jitter and packet loss
It is expected that many critical services which require ultra low latency, jitter and packet loss will be carried
by the 5G network. For example, an E2E latency requirement of 1ms is being considered for applications like
tactile communication, AR and mission-critical IoT.

This stringent QoS requirement is a big challenge for existing transport network, and some innovative data
plane technologies are being considered, e.g. IEEE802.1 TSN, FlexE and Detnet. By employing techniques such
as zero congestion loss, pinned path, packet replication and duplication, Detnet aims to create layer 3
forwarding path with determined QoS, while IEEE 802.1 TSN and FlexE and technologies for deterministic
layer 2 transport.

169