Mpirical provide compelling, interactive learning
across a range of delivery options. Live Onsite, Live Online, or OnlineAnytime -- we have
a training course that is ideal for you. For a no-obligations chat about your training
requirements, contact us at mpirical.com. As with the Radio Access Network, the 5G Core network
is a new architecture and in this session, we're going to consider what that architecture actually
involves, starting off with PDU Sessions and Quality of Service flows, we will then consider
the actual architecture and key components, and as part of that, network functions virtualization and
also network slicing. So this is our high level 5G system. We've already considered our New Radio
and our next-generation Radio Access Network, so our key focus here is the 5G Core network.
And this is a more detailed representation of that core network. Now this is the reference
point representation where we see our various core network elements connected up with different
network reference points. Now the first thing that we're actually going to pick out from this
diagram is actually the user plane connectivity and that user plane connectivity in 5G is called
a PDU session -- a Protocol Data Unit session, and you can see it runs from the mobile or from
the device, through the GNB to the UPF (the User Plane Function) and then on to the data
network. So traditionally in LTE for example, it would be called the 'Packet Data Network,' in
5G it's the 'Data Network,' but this is going to be the internet or perhaps the IMS or
a corporate LAN of some description. So no other devices in the network will be using
the connectivity associated with this particular PDU session. Now, in order to provide QoS within
our PDU session, we have what are called 'QoS flows'. Now, a QoS flow is effectively a flow
of user plane traffic which will be receiving a particular level of Quality of Service. Now
because we might have traffic with different QoS requirements, what that might mean is within the
PDU session there are several QoS flows actually in operation. To distinguish between them, each
QoS flow has a QoS flow ID. And to begin with, we will have, if you like, a default QoS flow. And
this default QoS flow, it will have a particular level of Quality of Service, but it won't have any
kind of packet filtering on it, so all packets can potentially go down this QoS flow if necessary,
however, as soon as we start adding additional QoS flows with different QoS requirements and
different QoS levels, then we will start filtering which traffic can actually go down these bearers.
So you can see that in actual fact, within a PDU session we could have several QoS flows in
existence and you can see that the kind of QoS flow characteristics we see for Quality of Service
are things like latency, priority, and whether or not this particular QoS flow is guaranteed
bit rate or non-guaranteed bit rate. And, as appropriate, these flows can be transient, so we
can actually remove flows if necessary. Now, being realistic about this, if the data network was the
internet for example, then we typically only need one QoS flow -- that would be the default flow and
that would be a best-effort QoS flow. If however, this is maybe 5G voice services and the data
network is the IMS, then we could have a QoS flow that's carrying the signalling associated
with voice, and then we'd have a separate QoS flow that's carrying the actual voice packets
themselves. And as traffic traverses the QoS flow, it actually has a QoS flow ID added to the traffic
protocol stack so we know exactly which floors a particular piece of data should go down. But
certainly, to have lots and lots of QoS flows, well, we really need to consider what scenario
would actually require that. As we move on to our actual architecture though, we'll look at
each of these various elements in turn now, all fundamentally designed to keep that PDU
session active for the subscriber and ensure that fundamentally, the PDU session follows them
as the subscriber moves around the network. So to begin with, the first element that we're going to
consider that we find in the core network, is the 'Core Access and Mobility Management Function'.
And this has got a similar role to the MME in LTE whereby it looks after mobility management.
So, in terms of subscriber mobility, we consider the fact that this AMF will always know either
the tracking area that the subscriber is in, or the potential cell that they're in, and it
really depends on whether the subscriber is idle or connected, respectively. The AMF though, also
plays a key role in security and registration, so it's the AMF that will be liaising with various
other subscriber databases to ensure that the subscriber is allowed on the network in the first
instance, and the AMF will play a key role in authenticating that subscriber within the network.
Finally, the AMF will also provide the device with a temporary identity which you can use whenever
it signals the network. That temporary ID is also used in paging as well. So that's the Core Access
and Mobility Management Function. Moving back to our architecture, the next element is the Session
Management Function. So traditionally, in LTE, it would be the MME that does mobility management
and session management. We can see that in 5G, that functionality has actually been split,
so it's the AMF that does mobility management and here, the Session Management Function does
session management. So what we're effectively talking about here is the establishment and the
modification and the teardown of our PDU sessions, so the SMF is directly involved in that, and
as part of that, it will be routinely liaising with the Policy Control Function to determine
whether or not a particular user data session is allowed to go ahead. Not only that, the
Session Management Function is directly involved with the establishment of the actual PDU
session connectivity. So that pipe that we draw on our diagram which is representative of the
PDU session, is actually a series of separate connections that must be set up in the network.
Now, those connections run through the User Plane Function and it's the job of the SMF to choose
which UPF. And also, if the data session is IP based, the SMF will also be allocating an IPV4 or
an IPV6 address. Now, I did say 'if' because PDU sessions can actually be based on purely ethernet
or even unstructured data in 5G. It's not all about IP like it was in LTE. As we move through
the network, we've got our User Plane Function, we've already started to touch on this, and as
you can see, it is an anchor point for NG-RAN mobility. So as I move around the Radio Access
Network, I will move from one gNB to another, but the UPF will remain that anchor point into
the network. So the user plane connectivity will always be running from the gNB to our UPF in the
core network. And because the UPF is directly sat on the user plane, it's an ideal point to enforce
Quality of Service that's ensuring that the right data is sent down the correct QoS flow, and also
implementing policy as appropriate. And that could be throttling of data, for example. As we move
through the network, the next element that we have is the UDM and this is Unified Data Management.
In effect, it's a central repository of subscriber information directly involved with Access
Authorization because it will be holding security keys. It's also involved in Registration and
Mobility Management because it will be tracking where our subscriber is attached to in terms
of which AMF our subscriber's been allocated, and then finally, it will contain the Data Network
Profile or profiles. It effectively contains the subscriber profile, telling entities like the
AMF and the SMF, exactly what our subscriber is and is not allowed to do, which data networks
they can connect to, and what kind of QoS profile they can expect to be granted when they do connect
to those data networks. Next in line and finally, is the PCF. And the PCF is there to implement
policy control, or 'Policy Control Function,' and when we talk about the PCF and implementing
policy control, it's all on a dynamic basis. Now, these dynamic policy decisions are based on
conditions that might be active in the network at this time. So before we just blindly set
up a PDU session to a particular data network, the SMF for example, will check in with the policy
control function to determine if, at this present time, are there any network conditions that are
going to influence how our subscriber experiences their service? So it might be that the fact
there's subscribers in a particular geographical location and because they are in that particular
cell, for example, the PCF determines that the subscriber needs to be throttled at this time,
or maybe isn't even allowed to get PDU session connectivity. So the PCF on the dynamic basis
has got the ability to alter both mobility and session-related service aspects. So it does play a
big part in the overall ecosystem. Now notice when we go back to our main diagram, the PCF does have
connectivity into the data network as well, so that PCF can take session-related information such
as our subscribers trying to make a phone call, it can take that information, send it into the 5G
Core network to ensure that the correct resources are established. Now, the title of this particular
slide is 'Network Functions Virtualization' because what is key for our 5G Core is that in
reality, much of these nodes will actually be virtualized as part of the NFV infrastructure. And
what do we mean when we say 'virtualize'? Well, effectively, these devices that we see are not
standalone devices. What they are are software processes running on what's termed a 'Commercial
Off The Shelf Server' and this is intrinsically, the Network Functions Virtualization architecture
whereby we see our virtualized devices -- whether it's control plane, potentially
user plane, subscriber management, billing and policy control -- they can all run as
software processes, maybe with a slight exception of the UPF -- there's an added added complexity
to the user plane, but certainly from a control plan perspective, these elements can run
as software processes and the idea behind network functions virtualization is you have a
Network Functions Virtualization infrastructure which is fundamentally there to provide these
software processes with the compute, the storage, and the network resources that they will all
undoubtedly, require, but the key facet is the NFV infrastructure is a shared infrastructure that
all of these software processes will actually use. And it's all built on Commercial Off The Shelf
hardware, so the cost savings, or the potential cost savings, to deploying the Core network based
on an NFV infrastructure are significant. Now, one of the key benefits of this, other than financial,
is the flexibility that you have in your network. Because these processes are running as software
processes, if we need to scale up or scale down capacity, it's much more straightforward
in a virtualized environment. So if we need more AMF capacity for example -- well, if it's a
traditional AMF deployed as a piece of hardware, then actually implementing a new AMF in the
network can take weeks or even months, whereas if it's a virtualized AMF, well, scaling up capacity
could be a matter of minutes, but crucially, you need what's called 'MANO' -- Management and
Orchestration to facilitate all of this. So this is a piece of the infrastructure in and of itself.
So fundamentally, this is actually what the 5G Core will look like; it's a series of virtualized
elements and what's crucial is traditionally, where we see protocols being exchanged between
our core network elements in networks like LTE and earlier technologies, in 5G, this is all API
driven: Application Programming Interfaces. These virtualized devices -- these virtualized elements
-- are sending API calls to one another in order to communicate. Now, the final consideration for
this session which is actually closely related, or may be an enabler for network slicing. So NFV is
a key enabler for this notion of network slicing, but what exactly is network slicing? Well,
network slicing effectively allows the service provider to create, if you like logical networks
across a common physical infrastructure. Now, if you recall, 5G is not just about providing huge
data rates to the subscriber. 5G is about becoming an enabler network for lots and lots of different
applications and lots and lots of different third-party users of the network. And the common
example is the Internet of Things. So you need to be able to create a very adaptive, flexible
network that will provide different customers different third parties with different features.
And network slicing is a perfect solution to do this because we can create these logical networks
across the same 5G physical infrastructure, so we ultimately provide that flexibility to
accommodate lots and lots of different potential service environments. So to illustrate, there's
our 5G RAN in our 5G Core network. We might have for example, a V2X Network slice and clearly
the characteristics of V2X are low latency, critical communication. Now on top
of the same physical infrastructure, we could have voice services for our typical
human subscribers. We need low latency for our voice and signaling and we need guaranteed Quality
of Service. And finally, we might have some kind of smart metering network slice where we need to
accommodate large numbers of connected devices, but they have actually fairly low data
requirements and relatively minimal network connectivity and activity potentially, so lots
and lots of different environments here that can be accommodated by network slicing. And
in terms of the standards, in actual fact, an individual device is able to connect to up to 8
network slices simultaneously. So in summary then, we talked to begin with, about PDU sessions
and we saw that they provide connectivity between the device and the data network.
Within a PDU session, we have QoS flows, and they are designed to provide specific levels
of Quality of Service for specific types of data running down them. We talked about the 5G Core
network, we talked about five of the key elements, namely the Core Access and Mobility Management
Function, the Session Management Function, the User Plane Function, Unified Data Management,
and then finally, the Policy Control Function. We talked about how all of these elements will
fundamentally be virtualized and a key benefit to virtualization is really opening the network
up to network slicing. Need to know more? Why not visit our store where you can choose from over
200 hours of video based training. Alternatively, you can contact us to discuss any specific
training requirements you may have.