Maybe you have already heard about some features as Dynamical Network Slicing, CloudRAN, Network-as-a-Services, … Some basic 5G principals we’ll briefly discussed also in this post. However my question is: What will be the change from the real-time communication point of view? What will be the 5G calling (VoNR) look like? Is the IMS (IP Multimedia Services, don’t confuse with International Microwave Symposium) to stay in the operators’ networks?
Seems that at the first stage the change will be less dramatic than when we introduced 4G. 4G was in many ways a revolution, whereas 5G is “only” an evolution. In fact 4G and 5G, at least in the beginning, will coexist and complement one each other. Still 5G will have a big impact on our existing technologies and the way we work with telecommunication networks.
It has been 3 years since I started this blog. It was around that time back in 2014 when I heard about 5G for the first time. It was funny, we were just doing the first 4G/VoLTE deployments, everything was new to operators and someone was already dreaming about the next version 🙂 In 2017 the 5G is still not finalized but we have done quite a few steps forward and 3GPP has created a lot of specifications (Rel-15) which uncover the concept (actually we have 5GPP and 5G.IEEE too!).
Let’s firstly define what we understand as 5G. Obviously, as the previous two generations, also the 5G brings more and faster data. This time we talk about “always-on connectivity”.
5G New Radio Scalable Network (5G NR) follows three main directions:
- enhanced Mobile BroadBand (eMBB)
- Enhanced Throughput
- Extended Coverage
- Ultra Reliable Low Latency Communication (URLLC)
- Lower Latency
- Location precision
- massive Machine Type Communication (mMTC)/ massive IoT
- Higher connection density
- Energy efficiency
To achieve faster data transmission the 5G defines NG-RAN in TS 38.804. A great article which summarizes 5G RAN Network Architecture can be found on Techplayon.
5G defines also new 5G Core Network. The 5G System architecture consists of the following network functions (NF):
- Authentication Server Function (AUSF)
- Access and Mobility Management Function (AMF)
- Network Slice Selection Function (NSSF)
- Policy Control Function (PCF)
- Session Management Function (SMF)
- Unified Data Management (UDM), Unified Data Repository (UDR)
- User plane Function (UPF)
- 5G-Equipment Identity Register (5G-EIR), NF Repository Function (NRF), Network Exposure Function (NEF), Unstructured Data Storage Function (UDSF), …
Well, that’s not the only or even the most significant change. In order to have network faster we don’t need to have a new Core Network infrastructure, do we? As Rick Hornby, executive director for core network planning in Verizon, put it: “Software defined Networking (SDN) and Network Functions Virtualization (NFV) are really a foundation for how 5G will be deployed. The separation of the control and user plane are all coming from what we see in the web-based environment and are part of what is happening with 5G.”
Many operators and telco vendors are trying to virtualize the existing 4G ecosystem. 5G has the virtualization harcoded directly in its DNA. Simply put with NFV we want to separate the network and the data services, so we can treat network as another resource which can be dynamically assigned as needed. This is very important because operators want to differentiate various classes of traffic data (Sensors NW, Smart City cameras, Augmented Reality, Self Driving Cars, Web Services, Real-time communication, …) and assign resources based on business priorities and SLAs. In the same way Amazon, Microsoft, Google, Atlantic and others do offer Cloud services, Mobile Operators are about to offer the connectivity.
SDN then goes hand in hand. NFV brings the resources, SDN the way how to manage and automate them. Currently most of the configuration of T1 networks is done statically. Usually the only dynamic configuration is provided by DNS. But with the rapid increase of connected devices and supported functionalities it is more and more difficult for architects, designers and engineers to design and configure data networks manually. SDN separates the (virtualized) network infrastructure from its logical architecture. It allows to automate configuration (e.g. routing, security) and management (e.g. scaling, security) procedures, and to control the Logical Network from one central place. We can say that SDN represents an abstraction layer above the (virtualized) network infrastructure.
Network Slicing is a way which allows mobile operators to partition a single network into multiple virtual networks. This makes possible for the operators to support many different types of services and/or multiple customers (service providers) using one physical infrastructure. 5GPP defines Network Slice as a composition of adequately configured network functions, network applications, and the underlying cloud infrastructure (physical, virtual or even emulated resources, RAN resources etc.), that are bundled together to meet the requirements of a specific use case, e.g., bandwidth, latency, processing, and resiliency, coupled with a business purpose. Network Slicing is an end-to-end concept covering all network segments including radio networks, wire access, core, transport and edge networks.
NFV defines how to virtualize the network functions, SDN the management, and Network Slicing bundles it together as a Network Service. Two main network slicing services can be considered:
Horizontal Slicing (Infrastructure Sharing)
The provisioning of Virtual Infrastructures (VI) under the control and operation of different tenants – in line with an Infrastructure-as-a-Service (IaaS) model, i.e., creation of a Network Slice Instance;
Vertical Slicing (QoS Slicing)
The provisioning of tenant’s owned Network Services (NS) as defined by the ETSI NFV architecture, i.e., creation of a Service Instance.
Compare to current situation in 4G when we often struggle to fully support multi-tenancy environment and when some operators have 2 or even more IMS networks supporting different services, e.g. VoLTE IMS, RCS IMS, IMS for Fixed networks, MVNO’s IMS… In 5G it should be easy to dynamically create a new virtual fully functional network dedicated to a particular tenant or service – a network slice. E.g. we can create a virtual 5GS network for New York City traffic cameras. To do that we don’t need any new physical infrastructure.
Service Based Architecture
Although we can still work with 5GS in the old-style as with Reference-point based system, our aim is to see the network from a new perspective. Virtualization and Slicing allows us to defined so-called Service Based Architecture (SBA), which remains backward compatible, but which is at the same time open to 3rd party applications and APIs, which can be running in a cloud as a Service. The basic principles followed by SBA 5G Core are:
- Control Plane and User Plane Separation (CUPS)
- scalability, redundancy, virtualization
- Concurrent Access to Services
- low latency (in local datacentres)
- e.g. central cloud-based Control Plane, local Media Plane Sites
- Decoupling of stateless processor and statefull backend (database)
- scalability, redundancy, virtualization
- Modular Function Design
- multiple instances of one network function
- Access Network (AN) and Core Network (CN) Separation
- Unified Authentication Framework
- Follow-the-user services
- Network Capability exposure
Is it new? Not really, a lot of these design patters we can see (for years!) already in 4G networks. As said in the beginning, 5G is many ways an evolution of 4G networks.
So what 5G brings? Till now we had more or less two separate worlds. A world of IT and Internet on one side. And a world of telco and mobile devices on the other. With mobile internet, with html5, with wifi, lte modems, and foremost M2M these worlds started to overlap. 5G is trying to be the new infrastructure for internet. The message is clear – get rid of cables, 5G can transmit all your data much faster wherever you are.
The limit is the sky The limit is the battery now. Connectivity and mobility becomes a default requirement and bandwidth is just another resource. Sure there is still a long way to go ..
And last but not least 5G again rises the question of Net Neutrality. A nice summary in this regard is provided by IEEE here. (Btw. It’s not just US and EU where the Net Neutrality is questioned – more here).
5G Calling and IMS
In the latest release of 3GPP 23.228 in Annex Y we can find the description of IMS in context of 5G System (5GS). 5GS is a 3GPP system consisting of 5G Access Network (AN), 5G Core Network and UE.
Other Impacted existing TS/TR
|TS/TR No.||Description of change||Target completion plenary#|
|24.229||SIP protocol and IMS procedures impacts of new type of IP-CAN;||CT#80 (June, 2018)|
|24.167||IMS MO impacts of new type of IP-CAN||CT#80 (June, 2018)|
|23.008||Potential impacts on subscription data related to IMS||CT#80 (June, 2018)|
|29.228||Potential impacts on Cx interface due to 5G||CT#80 (June, 2018)|
|29.229||Potential impacts on Cx interface due to 5G||CT#80 (June, 2018)|
|29.328||Impacts on Sh interface due to 5G||CT#80 (June, 2018)|
|29.329||Impacts on Sh interface due to 5G||CT#80 (June, 2018)|
|23.003||Potential impacts on numbering and addressing in IMS due to 5G||CT#80 (June, 2018)|
In theory there shouldn’t be many changes, as IMS is access independent. And indeed till now there are only a few things which have to be redefined or specified in more detail.
When it comes to IMS architecture, it stays as it is. The exception is HSS which now can be co-located or implemented as a part of the UDM. However the functionality of HSS should remain standalone.
The interfaces between IMS and 5GS are Gm and Rx. The same interfaces are used, i.e. Gm, from P-CSCF perspective towards the UE, when UE is in 5G System or any other IP-CAN and from IMS perspective, the Rx interface is used between the P‑CSCF and the Policy Control Function (PCF).
For communication with the IMS, the UE shall acquire an IP address. SMF shall support the capability to send the P-CSCF address(es) to UE. The IP address will be either an IPv4 address or and IPv6 address. If the P-CSCF and IMS-AGW do not support both versions, then network design is expected to ensure that IP address incompatibility does not occur.
The 5GC elements and UE support the following mechanisms:
- During PDU Session establishment procedure, the SMF sends the IP address to the UE via SM NAS signalling. The IPv4 address allocation and/or IPv4 parameter configuration via DHCPv4 can also be used once PDU Session is established.
- /64 IPv6 prefix allocation shall be supported via IPv6 Stateless Autoconfiguration according to RFC 4862. IPv6 parameter configuration via Stateless DHCPv6 (according to RFC 3736) may also be supported.
In order to support DHCP based IP address configuration, the SMF shall act as the DHCP server towards the UE for both HPLMN assigned dynamic and static IP addressing and for VPLMN assigned dynamic IP addressing. The 5GC may also support the allocation of a static IPv4 address and/or a static IPv6 prefix based on subscription information in the UDM or based on the configuration on a per-subscriber.
If the UE changes its IP address e.g. due to changes triggered by the 5GS procedures, then the UE needs to re-register in the IMS.
If the UE acquires an additional IP address, then the UE may perform an IMS registration using this additional IP address as the Contact address. If IMS registration is performed, this IMS registration may co-exist with the previous IMS registration from this UE and the UE shall be notified that this IMS registration results in multiple simultaneous registrations.
The P-CSCF address(es) is sent to UE by SMF.
Access Domain Selection
For UE originated calls, the 5GC capable UE performs Access Domain Selection (ADS). The UE firstly attempts initial registration in 5GC. If the UE is set to “voice centric” for 5GS (so not “data centric”) it always tries to ensure that Voice service is available. If not, then by disabling capabilities to access 5GS, the UE re-selects to E‑UTRAN first (if available). When the UE selects E-UTRAN, it continues with Voice Domain Selection (TS 23.221). Note, that voice centric UE, which is capable of both 5GC and EPC, and which is unable to obtain voice service in 5GS, shouldn’t select a cell connected only to 5GC (see the deployment options for 5GS on Techplayon).
When call termination is requested by IMS (SCC AS?), the UDM/HSS shall be able to query the serving AMF for T-ADS related information.
The AMF responds to the query with the following information unless the UE is detached:
- whether or not IMS voice over PS Session is supported in the registration area (s) where the UE is currently registered;
- the time of the last radio contact with the UE; and
- the current Access Type and RAT type.
There should be no NAT (or its existence should be kept transparent towards the UE) located between the UPF and the P‑CSCF, which is possible as they are either located within the same private network and share same address space, or both the UE and the P‑CSCF are assigned globally unique IP addresses.
All associated media flows (such as e.g. RTP / RTCP flows) used by the UE to support a single media component are assumed to be carried within the same 5GS QoS flow. The 5G QoS model supports both QoS flows that require guaranteed flow bit rate and QoS flows that do not require guaranteed flow bit rate. The 5G QoS model also supports reflective QoS. Within the 5GS, QoS flows are controlled by the SMF.
A network which requires the Geographical Identifier to be generated in the IMS may implement a mapping table between a NG-RAN cell identifier received as part of Access Network Information and a Geographical Identifier. The P-CSCF or an IMS AS may then, based on operator policy, use this mapping table to convert the user location into a Geographical Identifier, and insert the Geographical Identifier in the SIP signalling, thus enabling routing decision in downstream IMS entities or interconnected network.
Without IMS-level roaming interfaces
In this roaming model, which corresponds with LTE S8HR, the UPF holding the UE’s IP point of presence is located in the home PLMN and therefore UE IMS signalling and user plane are routed to home PLMN.
Architecture with IMS-level roaming interfaces
This model corresponds with Local Breakout (LBO) as we know from LTE.
In IMS LBO scenarios where P-CSCF is located in VPLMN, the home network determines the serving PLMN of the UE from the location of the P-CSCF during initial IMS Registration, using the P‑CSCF network identifier.
To summarize, 5G brings a lot of new. But when it comes Voice/Video calling, the impact in the beginning should be minimal. (And that’s a good news 🙂 ) But soon or later we can see NFV and SDN or even Network Slicing implemented also for IMS services – IMSasS.
There were some hopes, that with the new release IMS will be simpler (and therefore cheaper to implement and deploy). However right now it seems that the general IMS architecture remains untouched.
Well, maybe one update, it’s possible that the Voice service for 5G will be called VoNR.