5G Core (5GC) – Platform Architecture

The foundational difference between 4G and 5G architecture is that 5G incorporates cloud native technology based on microservices & containers as well as enabling technology such as NFV and SDN. The reason for this is to accommodate the explosion of use cases that 5G is intended to serve as explained here –  https://www.vamsitalkstech.com/cloud/the-three-key-use-case-areas-for-5g/

3GPP Goals for the 5GC

I have covered the main architectural principles of the 5G 3GPP design here – https://www.vamsitalkstech.com/architecture/the-four-key-5g-architecture-principles/

Based on these requirements, 3GPP has kept the overall architecture kind of relatively flat where both CP (Control Plane) and UP (User Plane) are separated in order to enable them to scale independently based on traffic requirements. The common AN-CN interface also enables integration and interoperability so that MNOs (Managed Network Operators) can design, dimension and moderate the network based on traffic patterns and use cases. 

While one runs into the term ‘mobile core’ in the specifications and in general industry documents & proceedings, it is just the 4G EPC (Evolved Packet Core) or NG-Core in 5G. And while the term Core seems to suggest that it is not really part of the Access Network, it actually is. The mobile core provides bridging access between the RAN which interfaces with User Equipment (UE) and the broader IP based internet. When imagined as an application composed of a set of microservices, each instance of the mobile core serves a good sized Metro area where the RAN would span hundreds or even lower numbers of cell towers serving a corresponding number of mobile subscribers. 

The Mobile Core does two things #1)  provides IP connectivity to the mobile customer while ensuring that users are authenticated and are receiving service that satisfies their promised SLA. What accounts for the architectural complexity you will see below is that users are moving all the time and the Core needs to keep track of them; #2) supports the massive scalability to provide mobile connectivity at low latency to IoT/Industrial Equipment etc (a key differentiators for 5G versus 4G) 

5GC Systems Architecture 

With the background out of the way, let us now discuss the main point of this blogpost, the architecture of 5G Core. As with the 5G NR architecture covered here – https://www.vamsitalkstech.com/5g/5g-new-radio-nr-platform-architecture/, the mobile core as well provides a range of functionality – 

1. It provides IP connectivity to the Access Network (and 5G NR functions) thus enabling both data and voice services 
2. Provides subscriber management functionality 
3. Ensures that promised QoS requirements to subscribers are met
4. Provides uninterrupted service as subscribers are mobile and roaming
5. Enables vendors to incorporate value added features such as cloud native orchestration, improved lifecycle management for 5G functions, rapid provisioning, 
6. Support co-existence with the 4G legacy network for years to come (via the Non Standalone Option – NSA) as we will discuss in subsequent blogs

The 5G Mobile Core follows a microservices architecture with its core functional blocks laid out as depicted in the below illustration. The Service Based Architecture (SBA) provides modular building blocks (both CP and UP) modeled as Network Functions (NFs) who invoke and use each other’s services in a way that different vendors . 

       Service Based 5G System Reference Architecture – Source: MetaSwitch

The 5G Mobile core can be divided into three main components based on CP and a corresponding equivalent in the EPC (Evolved Packet Core). 

The first group includes the below – 

1. AMF (Core Access and Mobility Management Function) – The AMF is analogous to the EPC’s MME from a mobility aspect. It is responsible for connection management, authentication/authorization, location services and reachability.  As depicted in the above architecture, the 5GC AMF receives all connection and session information from User Equipment (UE) and then passes on session related information to the SMF over the N11 interface while handling mobility related tasks. 
2. SMF (Session Management Function) – The SMF ,as the name suggests, is in charge of session management for User Equipment. It manages each UE PDU session, and performs a range of tasks such as  IP address allocation, selecting and routing to the relevant UP function, and sets the control aspects of QoS. The SMF corresponds to the MME & the control aspects of the PGW in EPC.  
3. PCF (Policy Control Function) – The PCF provides policy rules that are then carried out by control plane functions. These policies include rules for network slicing, mobility management. The PCF supports 5G QoS policy and charging control functions.  
4. AUSF (Auth Server Function) and UDM (Unified Data Management): As the name suggests, the AUSF authenticates users. The UDM is in charge of user identity management, generates authentication credentials and corresponds to the HSS portion of the EPCs HSS. Manages user identity, including the generation of authentication credentials. Both the AUSF and UDM make up the functionality in the EPC’s HSS.

As mentioned, the 3GPP is clearly and deliberately pointing to a microservices and cloud native design for both the 5GC as well as for RAN. The consistent theme across both is stateless design, open interfaces and supporting a high degree of scalability. That leads us to the next set of components which run in the Control Plane but don’t have corresponding elements in EPC.  

The second group includes – 

1. SDSF (Structured Data Storage Network Function) –  A service that is used to store structured data such as subscriber information as an example, in a way that microservices can access it. Implemented using a SQL database such as PostgreSQL. 
2. UDSF Structured Data Storage Network Function) –  A service that is used to store unstructured data such as device information, any other opaque data as an example. Microservices implementing other CP functions access this data. Implemented using a NoSQL database such as Redis. 
3. NEF (Network Exposure Function): The NEF corresponds to an API Server in a microservices architecture. This component is used to expose services or business capabilities to 3rd party systems.
4. NRF (Network Repository Function): A means of service discovery in the microservices based Control Plane.
5. NSSF (Network Slicing Selector Function): Network Slicing by itself deserves a series of blog posts. It provides the carrier to select, slice and deliver the three main categories of 5G usecases and capabilities (https://www.vamsitalkstech.com/cloud/the-three-key-use-case-areas-for-5g/) over the 5G network. The NSSF helps slice the Network to serve User Equipment (UE). 

The final building block of the 5G architecture is the UPF. 

The UPF (User Plane Function) connects the RAN and the IP internet. It also provides per-flow QoS handling, packet routing and forwarding, traffic usage reporting etc. The UPF corresponds to the S/PGW combination in EPC. 

Conclusion

We discussed the 5G Service Based Architecture in this post. CSPs know that 5G deployments will take both time and effort. To that end, both 4G and 5G implementations will coexist for years to come.  In subsequent posts we will discuss the 5G NSA (Non Standalone Architecture) as well as dive deeper into Containers and Service Mesh technology from both a 5GC and NR standpoint.