Most engineers who work with 5G networks can name the core network functions. Very few can explain what they do in operational terms what breaks when they fail, how they interact under load, and what it means to configure them correctly for a specific deployment. This article closes that gap.
Why the 5G Core Is Different From Everything Before It
The 5G Standalone core network represents a complete architectural departure from 4G EPC. It is not an evolution. It is a redesign from first principles, built around three ideas that previous generations did not embrace: cloud-nativeness, service-based architecture, and the separation of user plane from control plane.
In 4G, network functions were monolithic. The Serving Gateway handled both user data and control signaling. The PDN Gateway combined session management with data forwarding. Scaling one function meant scaling everything bundled with it. Upgrading one component risked disrupting functions it shared hardware with.
5G SA breaks all of this apart. Each network function has a single, well-defined responsibility. Functions communicate through standardized service-based interfaces using HTTP/2 and JSON, the same protocols that run the modern internet. Every function runs as a containerized workload, scalable independently, deployable on commodity hardware.
This architecture is more flexible, more efficient, and more powerful than 4G EPC. It is also more complex to understand and operate because the simplicity of the interfaces hides the depth of what each function actually does.
The AMF: Access and Mobility Management Function
The AMF is the first network function a device encounters when connecting to a 5G SA network. It is the entry point for all access and mobility signaling, and it maintains the connection between the device and the rest of the 5G core throughout the device’s time on the network.
What the AMF actually does
When a device powers on and sends a registration request, the AMF receives it. The AMF authenticates the device in coordination with the Unified Data Management function, which holds subscriber credentials. It authorizes the device’s access, assigns a temporary identity called the 5G-GUTI, and registers the device’s location in the network.
Throughout the device’s active session, the AMF tracks mobility. When a device moves between cells and triggers a handover, the AMF coordinates the process maintaining session continuity, updating location records, and ensuring that the device’s data path is preserved through the transition. When a device enters idle mode to save power, the AMF maintains enough context to page the device when incoming traffic arrives.
The AMF also handles Non-Access Stratum signaling the control messages between the device and the core that manage registration, authentication, security, and mobility. All NAS messages from the device are terminated at the AMF, regardless of which RAN node they arrive from.
What breaks when the AMF is misconfigured
AMF configuration errors surface as authentication failures, registration rejections, or mobility failures. A common misconfiguration is incorrect operator network identifiers if the AMF’s configured PLMN ID does not match the network’s broadcast PLMN, devices will fail to register even with valid credentials. Incorrect security algorithm configuration produces authentication failures that look identical to credential problems until you trace the NAS signaling in detail.
AMF capacity misconfigurations produce subtler problems. Under high registration load during network recovery after an outage, for example an underdimensioned AMF will drop registrations or increase latency to the point where devices time out and retry, compounding the load. Engineers who understand AMF dimensioning can recognize this pattern and respond correctly. Engineers who don’t tend to escalate to vendors for problems that are operationally solvable.
For teams building operational competency in 5G SA core functions, the 5G training programs at 5GWorldPro cover AMF operations with hands-on lab scenarios specifically designed around failure modes and recovery procedures not just happy-path configuration.
The SMF: Session Management Function
The SMF manages the lifecycle of PDU sessions the data connections between devices and external networks. Where the AMF handles who is on the network and where they are, the SMF handles what data connections they have and how those connections are configured.
What the SMF actually does
When a device requests a data connection, the request flows from the AMF to the SMF. The SMF selects a UPF instance to serve the session, allocates an IP address from the appropriate pool, applies the QoS policies that correspond to the device’s subscription and the requested Data Network Name, and instructs the UPF to establish the forwarding rules for the session.
Throughout the session, the SMF monitors and enforces policy. It communicates with the PCF the Policy Control Function to receive and apply policies governing traffic handling, QoS parameters, and charging rules. If a policy change is triggered mid-session, the SMF updates the UPF’s forwarding behavior without interrupting the data flow.
Download the Medium app
The SMF also manages URSP rules UE Route Selection Policy which govern how multi-homed devices distribute traffic across multiple PDU sessions. For enterprise private 5G deployments where devices may simultaneously maintain sessions to the operator network and to local enterprise services, URSP configuration is a critical operational parameter that significantly affects application performance.
The SMF’s role in network slicing
Network slicing multiplies the SMF’s complexity. In a sliced deployment, different SMF instances serve different slices — each with its own QoS framework, charging logic, and UPF selection criteria. A single device connected to multiple slices simultaneously has separate PDU sessions managed by separate SMF instances with separate policy contexts.
Operational teams managing sliced deployments need to understand slice-specific SMF configuration in enough detail to troubleshoot cross-slice performance issues, diagnose session establishment failures that stem from slice selection errors, and validate that QoS policies are being correctly applied per slice. This is an operational skill that does not come from reading 3GPP specifications it requires hands-on practice with real sliced deployments, the kind covered in structured 5G training built specifically for core network operations.
The UPF: User Plane Function
The UPF is where user data actually flows. Everything else in the 5G SA core is control plane signaling, policy, management. The UPF is the data plane. Every packet to and from a connected device passes through a UPF instance.
What the UPF actually does
The UPF receives forwarding rules from the SMF through the N4 interface using the PFCP protocol. These rules tell the UPF exactly what to do with packets matching specific criteria: which traffic to forward to which destination, which QoS marking to apply, which packets to inspect for policy enforcement, and how to handle packets that don’t match any rule.
In practice, the UPF performs several distinct functions simultaneously. It terminates GTP-U tunnels from the RAN the encapsulation protocol that carries user plane data between the radio access network and the core. It enforces QoS policies on a per-flow basis, marking packets with DSCP values and applying traffic shaping to maintain the QoS commitments made by the SMF. It performs uplink classifier and branching point functions for multi-homing scenarios. And it collects usage reporting data that feeds into charging systems.
UPF placement and latency
UPF placement is one of the most consequential architectural decisions in a 5G SA deployment, and it is frequently under-appreciated until performance problems emerge.
A UPF placed centrally in an operator’s data center introduces round-trip latency that may be acceptable for consumer mobile broadband but is incompatible with industrial automation, remote operation, or other latency-sensitive enterprise use cases. The solution is to deploy UPF instances at the network edge co-located with or adjacent to the enterprise application servers so that user plane data takes the shortest possible path.
This is the foundation of Multi-Access Edge Computing. The UPF’s ULCL Uplink Classifier function allows the SMF to configure local breakout, where specific traffic flows are steered to a local UPF at the edge while other traffic continues to the central UPF for internet access. Configuring ULCL correctly for a specific enterprise deployment requires understanding both the UPF’s data plane functions and the SMF’s session management logic another intersection that demands structured training rather than trial and error on a live network.
How AMF, SMF, and UPF Work Together
The three functions operate as a coordinated system, but their coordination is mediated entirely through signaling they share no internal state and communicate only through their defined interfaces.
A device registration and session establishment sequence illustrates this clearly.
The device sends a registration request that arrives at the AMF via the RAN. The AMF authenticates the device through the UDM and AUSF, establishes a security context, and completes registration. The device then sends a PDU session establishment request this also arrives at the AMF, which selects an SMF and forwards the request.
The SMF receives the session request, selects a UPF, allocates an IP address, retrieves QoS policy from the PCF, and sends PFCP session establishment to the UPF with the forwarding rules for this session. The UPF acknowledges. The SMF sends a session response to the AMF with the session parameters. The AMF sends the session acceptance to the device through the RAN. The RAN establishes the data radio bearers. The device begins sending and receiving data through the UPF.
Each step in this sequence involves specific interface messages, specific parameters, and specific error conditions. Engineers who understand this sequence at an operational level not just conceptually, but in terms of what the message trace looks like, where failures surface, and how to diagnose specific error codes can troubleshoot session establishment problems in minutes rather than hours.
What This Means for Operations Teams
The 5G SA core is not difficult to understand once you have the right framework. But the operational knowledge required to run it well to configure functions correctly for specific use cases, to diagnose failures by reading interface traces, to dimension correctly for traffic forecasts, to manage slices without cross-slice interference is substantively different from what 4G operations required.
The engineers who develop this operational knowledge before a deployment goes live consistently outperform those who develop it after. Faster fault diagnosis, fewer vendor escalations, better network performance from day one. The investment is in training that is specific enough to actually transfer to operational behavior not general awareness programs, but curricula built around the actual functions, interfaces, and failure modes of the specific network being deployed.
5GWorldPro offers 5G SA Core training programs built around real operational scenarios AMF, SMF, UPF, PCF, NEF, and NSSF with hands-on lab environments and failure-mode exercises. Full curriculum at 5gworldpro.com/5g-training.
Top comments (0)