Evolved Packet Core (EPC): Architecture, Pros/Cons, and Best Practices

What is Evolved Packet Core (EPC)? 

The Evolved Packet Core (EPC) is the all-IP core network architecture defined by 3GPP for LTE and System Architecture Evolution (SAE). It provides the control and user-plane functions that support mobility, session management, policy enforcement, charging, and connectivity to external packet data networks. In practical deployments, EPC can also support interworking with earlier 3GPP access networks and, through defined mechanisms, selected non-3GPP access scenarios.

Key components of the EPC:

• Mobility management entity (MME): The primary control node, handling session management, authentication, and user tracking (mobility) across the network.
• Serving gateway (SGW): Routes and forwards user data packets while acting as the anchor point for mobility between different access technologies.
• Packet data network gateway (PGW): Connects the 4G network to external packet data networks (Internet) and manages IP address allocation, QoS enforcement, and charging.
• Home subscriber server (HSS): A central database containing subscriber-related information, user profiles, and authentication data.
• Policy and charging rules function (PCRF): Enforces service data flow policies and controls charging functionalities for users.

This is part of a series of articles about cellular technologies

History of Evolved Packet Core

EPC was first introduced with 3GPP Release 8 to support LTE. Early versions focused on separating the control and user planes to increase scalability and efficiency. Over time, new features were added to support advanced use cases.

In Release 10 and 11, EPC introduced features for heterogeneous networks (HetNets) and small cell support. These changes improved network coverage and capacity. Release 12 and 13 added support for device-to-device communication and enhanced machine-type communication (eMTC), addressing growing IoT demands.

The transition toward 5G started with Release 14 and 15, where EPC evolved into a more flexible and cloud-native architecture. While the 5G core (5GC) is a complete redesign, many operators initially deployed 5G radio with EPC in a non-standalone (NSA) configuration to accelerate rollout.

Today, EPC continues to coexist with 5GC in hybrid networks, especially in areas where full 5G standalone deployment is not yet available.

Related content: Read our guide to 5G SA

Key Architecture Components of the EPC

Mobility Management Entity (MME)

The MME is the control-plane node responsible for managing mobility, session setup, and signaling for user equipment (UE) in LTE networks. It handles initial UE attachment, bearer establishment, and tracking area updates. When a UE powers on and connects to the network, the MME performs the attach procedure, authenticates the user via the HSS, and initiates the creation of default and dedicated bearers.

The MME also manages handovers between eNodeBs that involve only signaling (S1-based handover) and coordinates with the SGW to update user plane routes. It supports paging for incoming data or calls and maintains idle mode UE contexts to reduce signaling overhead.

Security functions such as generation and distribution of encryption and integrity protection keys are also managed by the MME. It supports lawful interception, location services, and interacts with external systems like the EIR (Equipment Identity Register) to validate device identities.

Serving Gateway (SGW)

The SGW is a user-plane node that routes and forwards IP packets between the LTE access network and the core. It serves as the anchor point for intra-LTE mobility, ensuring uninterrupted data sessions as the UE moves across eNodeBs. During handover, it buffers downlink data and forwards it once the new path is established.

The SGW manages GTP-U tunnels for each bearer and updates tunnel endpoints as UE location changes. It also interfaces with the PGW for external data connectivity and may interact with legacy systems during inter-RAT (radio access technology) handovers, such as LTE-to-3G.

Other SGW responsibilities include generating charging data records (CDRs), supporting lawful interception, and optionally performing downlink packet marking based on QoS settings received from the MME and PGW. The SGW can also support localized breakout for edge data offloading in some deployment models.

Packet Data Network Gateway (PGW)

The PGW serves as the interface between the LTE network and external packet data networks such as the internet, IMS, or private enterprise networks. It is responsible for allocating IP addresses to UEs, managing policy enforcement, and applying QoS and charging rules as defined by the PCRF.

Each UE is associated with one PGW per PDN connection. The PGW performs deep packet inspection (DPI) to enforce application-level policies and supports per-flow QoS by mapping traffic to appropriate bearers. It also provides address translation (NAT) and firewall functionalities, especially in consumer internet scenarios.

The PGW enables seamless mobility across 3GPP (e.g., LTE to UMTS) and non-3GPP (e.g., Wi-Fi) access networks using PMIPv6 or GTP-based mobility protocols. In multi-operator environments, the PGW may enforce APN-based traffic segregation and support VPN tunnels for enterprise users.

Home Subscriber Server (HSS)

The HSS is a centralized database that contains user subscription information, including user identifiers (IMSI), authentication vectors, QoS profiles, access restrictions, and PDN configuration details. It supports multiple functions by interacting with core nodes using the Diameter protocol.

When a UE initiates a connection, the MME queries the HSS to retrieve authentication vectors using the Authentication and Key Agreement (AKA) protocol. The HSS returns information such as the default APN, allowed PDNs, and subscribed QoS parameters.

The HSS supports roaming by interfacing with other HSSs or using the SLF (Subscriber Location Function) to resolve the correct database in distributed deployments. In networks that support IMS, the HSS works with the IMS HSS or UDR (User Data Repository) to share subscriber data across voice and data services.

Policy and Charging Rules Function (PCRF)

The PCRF is a control-plane function responsible for real-time policy decision-making related to QoS, service prioritization, and charging. It interfaces with the PGW over the Gx interface and with online/offline charging systems (OCS/OFCS) to enforce data usage rules.

When a new bearer is created, the PGW queries the PCRF to receive policy rules based on the user profile, current network conditions, and application requirements. These rules define allowable bandwidth, traffic treatment (e.g., priority, delay tolerance), and charging parameters.

The PCRF enables support for service-based charging, zero-rating of specific applications, and dynamic changes to policies based on usage thresholds or service triggers. It plays a key role in enabling services like video streaming optimization, parental controls, and enterprise-level traffic shaping.

In VoLTE deployments, the PCRF is essential for managing voice bearer prioritization and supporting IMS-specific policies to ensure call quality and reliability.

EPC in LTE Networks

In LTE networks, the Evolved Packet Core (EPC) serves as the central backbone that manages both user data and control signaling. It supports all-IP connectivity, replacing the circuit-switched elements of earlier generations with a flat, packet-based architecture optimized for high-speed mobile broadband.

The EPC connects the LTE radio access network (eNodeBs) to external IP networks and provides key functions such as mobility management, session management, QoS enforcement, and IP address allocation. User devices (UEs) access the internet and other services through bearer paths established between the UE, eNodeB, SGW, and PGW.

During initial attach, the UE communicates with the MME, which handles authentication via the HSS and coordinates the setup of bearer channels through the SGW and PGW. Once established, the user plane traffic flows directly between the UE and the PGW through the SGW, bypassing the MME.

EPC also enables features such as seamless handover between eNodeBs, interworking with legacy 3G networks, lawful interception, and support for voice services through the IMS. Its modular design allows network operators to scale capacity, manage subscriber policies in real time, and evolve toward 5G through software-defined and virtualized deployments.

Evolved Packet Core Benefits and Challenges

Evolved Packet Core (EPC) introduced a streamlined, all-IP architecture that significantly improved mobile network performance compared to earlier generations. However, despite its advantages, EPC also presents several technical and operational challenges.

Pros:

• All-IP Architecture: EPC enables efficient data transport and simplifies integration with IP-based services, improving scalability and reducing legacy dependencies.
• Flat Network Design: Eliminates intermediate nodes found in older architectures, resulting in lower latency and faster data throughput.
• Separation of Control and User Planes: Improves flexibility and supports more scalable and efficient resource allocation across different network functions.
• Enhanced QoS Management: Enables fine-grained control over traffic prioritization and bandwidth allocation, supporting services with diverse performance needs.
• Support for Seamless Mobility: Maintains session continuity during intra-LTE and inter-RAT handovers, which is critical for uninterrupted voice and data services.
• Integration with Policy and Charging Systems: Real-time enforcement of user-specific policies enables dynamic service differentiation and flexible billing models.
• Readiness for Virtualization: Many EPC components can be virtualized, enabling deployment in NFV-based environments and supporting network slicing in transition to 5G.

Cons:

• Complex Signaling Procedures: Control-plane signaling across MME, SGW, and PGW introduces complexity, especially under high user mobility or dense network conditions.
• Scalability Constraints: Although more scalable than legacy cores, centralized EPC components like the PGW can become bottlenecks in high-traffic networks without proper load balancing.
• Interoperability Challenges: Integrating EPC with legacy 2G/3G systems and non-3GPP access networks can be complex due to protocol and architecture differences.
• Security Risks: Being fully IP-based increases exposure to IP-layer attacks, requiring robust security mechanisms across all interfaces.
• Transition Complexity: Migration from EPC to 5G Core (5GC) requires careful planning, especially in hybrid deployments with both NSA and SA architectures.

Latency in NSA Deployments: Using EPC in 5G NSA mode introduces additional latency compared to pure 5GC-based SA configurations, limiting some advanced 5G use cases.

Evolved Packet Core Design Best Practices

Adopt Control-User Plane Separation (CUPS)

CUPS allows operators to deploy control plane (SGW-C, PGW-C) and user plane (SGW-U, PGW-U) functions independently. This supports flexible scaling, where user plane capacity can grow with data traffic demands without increasing control plane overhead. It also allows user plane nodes to be placed closer to the edge of the network, reducing round-trip time and supporting latency-sensitive applications such as VoIP or mobile gaming.

Control plane functions can remain in centralized data centers for simplified signaling management, policy control, and coordination. This separation enables more efficient utilization of compute resources, especially in distributed deployments. CUPS also facilitates easier integration with MEC (Multi-access Edge Computing) and improves support for emerging 5G-like services in LTE networks.

Operationally, CUPS supports rapid fault isolation and recovery by limiting the impact of control or user plane node failures to only their respective functions. It also simplifies upgrades, enabling independent lifecycle management of control and user plane elements.

Design for Virtualization and Cloud-Native Deployment

Virtualizing EPC components enables network functions to run on commodity hardware in data centers, reducing CAPEX and OPEX. Operators can scale functions like MME or PGW on demand based on load, improving resource efficiency and energy use. Functions are deployed as VNFs in traditional NFV environments or as CNFs in modern containerized infrastructure.

Cloud-native EPC design takes this further by enabling stateless microservices, API-based communication, and container orchestration. This improves service agility, allowing continuous integration and deployment (CI/CD) and rapid delivery of new features. Cloud-native EPC functions can scale horizontally, recover automatically, and be deployed across hybrid cloud environments.

These architectures also support zero-touch provisioning and full lifecycle automation through orchestration frameworks like ETSI MANO or Kubernetes-based platforms. Cloud-native principles lay the foundation for network slicing, automated fault recovery, and 5G migration readiness.

Efficient Resource Allocation

EPC performance depends on balanced distribution of control and user plane workloads. For example, MME load can spike due to signaling storms or frequent mobility events, while PGW can become saturated with heavy data traffic. Resource planning should consider control plane metrics (e.g., session setup rate, attach success rate) and user plane throughput (e.g., Mbps per PGW-U).

Dynamic scaling policies can be implemented using network analytics and monitoring tools that trigger instantiation of new VNFs or containers. Load balancers should distribute traffic across multiple SGWs and PGWs to avoid single points of congestion. High-availability clusters and horizontal scaling help meet service-level agreements (SLAs) even during peak periods.

Network slicing and QoS enforcement can be used to allocate bandwidth and processing capacity per service or subscriber group. This allows operators to isolate traffic from enterprise users, IoT devices, and consumers, preventing resource contention across services.

Secure APIs and Signaling Protocols

EPC interfaces, including S1-MME, S11, S5/S8, Gx, and Diameter-based signaling, are critical attack surfaces. These interfaces must be protected against spoofing, replay attacks, denial-of-service (DoS), and message tampering. Using IPsec tunnels and mutual authentication between nodes helps secure control and user plane communication.

Diameter interfaces must be hardened with message validation, overload protection, and origin host verification to prevent malicious signaling attacks. Signaling firewalls can detect malformed or unexpected messages and enforce protocol compliance. TLS should be used to encrypt API calls, especially when exposing northbound interfaces to external systems.

Network functions should also implement access control and rate-limiting to prevent abuse. In multi-tenant environments, strict isolation between virtual EPC instances ensures one tenant’s traffic or misconfigurations don’t affect others. Continuous security monitoring and adherence to 3GPP security standards (e.g., TS 33.401) are essential for protecting subscriber data and network integrity.

Plan for Smooth Evolution to 5G Core

EPC must be designed with forward compatibility to reduce disruption during the shift to 5G. Deploying CUPS and virtualized network functions early allows reuse of user plane elements in 5GC architectures. This is especially useful in 5G NSA deployments, where 5G NR radios connect to the existing EPC instead of a new 5GC.

Operators should also ensure interoperability with 5GC components using standardized interfaces such as N26 (interworking between MME and AMF) and NG interconnects. EPC user plane nodes can serve both LTE and 5G traffic if appropriately designed, enabling shared infrastructure and smoother transition.

Network orchestration and automation platforms used in virtual EPC deployments should support multi-domain service chaining and network slicing, as required in 5GC. Planning should also address migration of subscriber data and policy rules from HSS and PCRF to the UDM and PCF functions in 5GC.

A gradual migration strategy, starting with edge upgrades and integrating 5G NR in NSA mode, can provide service continuity while modernizing the core. This reduces risk, minimizes CAPEX spikes, and supports operator-specific timelines for full 5G standalone deployment.

Deploying 5G for IoT Connectivity with floLIVE

For organizations running international IoT deployments, the packet core discussion is not just theoretical. Breakout location, packet gateway placement, and policy control all affect latency, resilience, and data residency.

floLIVE positions its offering around a cloud-native, distributed core network and local breakout model for IoT. That approach helps enterprises, MNOs, and MVNOs keep traffic closer to the application or region where it is used, rather than backhauling everything through a distant home network.

That positioning is especially relevant for teams supporting LTE-based IoT fleets today while preparing for broader 5G adoption. In those cases, localized packet gateway deployment and unified policy control can help improve performance and support regional data handling requirements across markets (like GDPR).