Serving Gateway User plane function

Description

The Serving Gateway User plane function (SGW-U) is the packet-forwarding engine resulting from the disaggregation of the traditional Serving Gateway. Introduced with Control and User Plane Separation (CUPS), the SGW-U handles all the data plane processing for user traffic in the Evolved Packet Core (EPC). Its primary role is to serve as the mobility anchor for the user plane during inter-eNodeB handovers within LTE and to route and forward traffic between the Radio Access Network (RAN) and the Packet Data Network Gateway (PGW). It operates under the direct command of its associated control plane function, the SGW-C, using the Packet Forwarding Control Protocol (PFCP) over the Sxa reference point.

Functionally, the SGW-U performs packet inspection, classification, and enforcement actions based on rules installed by the SGW-C. These rules are conveyed through PFCP messages and include Packet Detection Rules (PDRs), Forwarding Action Rules (FARs), and QoS Enforcement Rules (QERs). A PDR defines how to identify a packet flow (using parameters like source/destination IP, port, and GTP tunnel endpoint identifiers). Once a packet matches a PDR, the associated FAR dictates the action, such as forwarding the packet to a specific next-hop tunnel (e.g., towards an eNodeB or a PGW-U), buffering, or dropping it. QERs allow the SGW-U to apply rate policing, marking, and scheduling to ensure the agreed-upon Quality of Service. The SGW-U is also responsible for encapsulating and decapsulating user IP packets within GPRS Tunneling Protocol (GTP-U) tunnels that connect it to the eNodeB (S1-U) and the PGW-U (S5/S8-U).

Architecturally, the separation allows the SGW-U to be implemented as a high-performance, streamlined packet processing node, often using hardware acceleration. It can be deployed in a distributed manner, much closer to the RAN than a centralized data center. This distribution is critical for reducing latency, which benefits applications like mobile gaming, autonomous vehicles, and augmented reality. In 5G interworking scenarios, the SGW-U can be collocated with or interconnected to a 5G User Plane Function (UPF) to facilitate seamless user plane handovers between 4G and 5G access. Its specifications span architecture (23.214), management (28.708, 32.867), the PFCP protocol (29.244), and security requirements (33.127).

Purpose & Motivation

The SGW-U was developed to overcome the inflexibility of the integrated SGW appliance. In traditional EPC deployments, the SGW's user plane was tied to its control plane, preventing independent optimization and scaling. This coupling was ill-suited for emerging trends like network function virtualization (NFV) and the need for low-latency edge computing. Scaling user plane throughput required provisioning entire new SGW instances, including redundant control plane resources, leading to inefficient capital expenditure.

The creation of the SGW-U, finalized in 3GPP Release 14 as part of CUPS, was driven by the desire for a more agile, cloud-native network architecture. By extracting the user plane into a separate function, operators can deploy lightweight, stateless SGW-U instances using commercial off-the-shelf hardware or virtualized platforms at optimal locations in the network topology. This enables traffic breakout at the edge for local services and reduces backhaul costs. The SGW-U's programmability via PFCP also opens the door for software-defined networking (SDN) principles in the mobile core, allowing for more dynamic traffic steering and service chaining. This separation was a direct precursor to the native control/user plane separation seen in the 5G Core's SMF and UPF, proving the concept for next-generation networks.

Key Features

• High-performance packet forwarding and GTP-U tunneling
• Enforces QoS policies (rate limiting, marking) via QERs
• Executes packet handling actions (forward, drop, buffer) via FARs
• Identifies user flows based on PDRs received from SGW-C
• Anchors user plane for intra-LTE handovers
• Supports distributed deployment at network edge for low latency

Evolution Across Releases

Defining Specifications