Internet Protocol Flow

Description

An IP Flow is defined in 3GPP specifications as a unidirectional flow of IP packets with the same source IP address and port number, the same destination IP address and port number, and the same transport protocol. Port numbers are only applicable if used by the transport protocol. This definition, standardized in 3GPP TS 23.228, TS 29.212, and TS 29.214, provides a precise, packet-level classification mechanism that enables network elements to identify and treat traffic belonging to a specific application or service.

In the 3GPP architecture, IP Flows are the building blocks for policy and charging control (PCC). The Policy and Charging Rules Function (PCRF) or, in 5G, the Policy Control Function (PCF), uses IP Flow information to install rules in the User Plane Function (UPF) or Packet Data Network Gateway (PGW). These rules, known as Packet Detection Rules (PDRs) or Traffic Flow Templates (TFTs), map IP Flows to bearers or QoS Flows. For example, a VoIP call might consist of multiple IP Flows: one for SIP signaling (e.g., UDP port 5060) and one or more for RTP media (dynamic ports). Each flow can be assigned a different QoS class, ensuring that voice packets receive higher priority than signaling or background data.

The concept of IP Flow is central to the Service Data Flow (SDF) concept in PCC. An SDF is a set of IP Flows that belong to the same service and are treated identically for policy and charging purposes. The SDF template, defined in TS 29.212, contains packet filters that identify the IP Flows. These filters are evaluated against each packet's header fields: source/destination IP address, source/destination port, protocol, and optionally other parameters like Type of Service (ToS) or Flow Label. When a packet matches a filter, it is classified into the corresponding SDF and receives the associated QoS and charging treatment.

IP Flows are also used in the context of bearer binding and QoS enforcement. In LTE, each EPS bearer can carry multiple IP Flows, but all flows within a bearer receive the same QoS treatment. In 5G, the QoS Flow concept allows finer granularity: each QoS Flow can carry one or more IP Flows, and the QoS Flow is the unit of QoS enforcement. The mapping of IP Flows to QoS Flows is performed by the UPF based on PDRs provided by the SMF. This mapping ensures that packets belonging to the same application receive consistent treatment across the network.

From a management perspective, IP Flow information is used for traffic detection, accounting, and charging. The Traffic Detection Function (TDF) can inspect IP Flows to identify application types and apply specific policies. Charging systems use IP Flow records to generate usage reports, enabling operators to bill subscribers based on the actual services consumed. The granularity of IP Flow classification allows operators to offer differentiated services, such as zero-rated traffic for specific applications or premium QoS for enterprise customers. Overall, the IP Flow concept is a cornerstone of 3GPP's ability to manage, prioritize, and monetize IP-based traffic in mobile networks.

Purpose & Motivation

The IP Flow concept was introduced to provide a standardized, granular way to identify and manage individual packet streams within a mobile network. Before its definition, traffic classification was often coarse-grained, based on bearer-level or APN-level parameters, which limited the ability to apply differentiated QoS and charging policies. As mobile networks evolved to support multiple simultaneous services (voice, video, data) over a single IP connection, the need for finer traffic differentiation became critical.

IP Flows solve this by enabling the network to distinguish between different applications or media streams based on packet header information. This allows operators to enforce QoS policies per flow, ensuring that latency-sensitive traffic (e.g., VoLTE) is prioritized over best-effort traffic (e.g., file downloads). It also enables flexible charging models, such as charging per application or zero-rating specific services. The concept was first standardized in 3GPP Release 7 as part of the Policy and Charging Control (PCC) framework, and it has since been extended and refined in subsequent releases to support new services and architectures, including 5G.

Historically, the IP Flow definition addressed the limitations of earlier approaches that relied on static bearer configurations. By providing a dynamic, packet-level classification mechanism, it allowed the network to adapt to changing traffic patterns and service requirements. This flexibility is essential for supporting the diverse range of applications in modern mobile networks, from real-time communication to IoT and cloud services.