Quality of Service

Description

Quality of Service (QoS) in 3GPP systems is a comprehensive framework for differentiating and prioritizing data traffic to meet the specific performance requirements of various services. It operates by classifying packets into distinct QoS Flows, each associated with a QoS Profile containing a set of QoS parameters. The core parameters are the 5G QoS Identifier (5QI), which is a scalar that references standardized characteristics (like priority, packet delay budget, packet error rate), and optionally Allocation and Retention Priority (ARP), Guaranteed Flow Bit Rate (GFBR), Maximum Flow Bit Rate (MFBR), and Averaging Window. These parameters define the expected treatment for packets belonging to that flow.

Architecturally, QoS is enforced end-to-end across the 5G system, involving the User Equipment (UE), the Radio Access Network (RAN), and the 5G Core Network (5GC). The Session Management Function (SMF) is responsible for establishing, modifying, and releasing QoS Flows based on session requests from the Application Function (AF) or subscriber profiles from the Unified Data Management (UDM). The SMF derives the QoS rules and sends them to the UE and the RAN via the Access and Mobility Management Function (AMF) and the gNB. In the RAN, these rules are mapped to Data Radio Bearers (DRBs) for over-the-air transmission, where scheduling algorithms prioritize packets according to their 5QI.

How it works involves a multi-layer mapping process. At the PDU Session level, service data flows (SDFs) are identified by packet filters and mapped to a QoS Flow. Each QoS Flow is uniquely identified by a QFI (QoS Flow Identifier) within a PDU Session. The RAN maps one or more QoS Flows with similar characteristics to a single DRB to optimize radio resource usage. The gNB's packet scheduler uses the QoS parameters (especially priority and delay budget) to make millisecond-level decisions on which UE's data to transmit on which physical resource blocks, ensuring that latency-sensitive flows like VoIP are served before background downloads. This hierarchical and granular control is what enables a single physical network to simultaneously support mission-critical IoT, ultra-HD video, and best-effort web browsing.

Purpose & Motivation

QoS technology was created to address the fundamental challenge of supporting multiple services with vastly different performance requirements over a single, shared packet-switched network infrastructure. Early cellular networks were circuit-switched and dedicated to voice, inherently providing guaranteed quality. With the migration to all-IP networks (GPRS, UMTS), all traffic became data packets, risking that latency-sensitive voice or video would be degraded by bulk data transfers. QoS provides the necessary tools to reintroduce differentiation and guarantees in a packet-based world.

The evolution from 3G to 4G and 5G saw a continuous refinement of QoS mechanisms to support an ever-broader range of services. In 3G/UMTS (Release 99), QoS introduced the concept of bearer services with traffic classes (Conversational, Streaming, Interactive, Background). 4G/LTE simplified this with QCI (QoS Class Identifier). 5G significantly enhanced the framework with a more flexible, service-based approach. It addressed limitations of previous systems by allowing more granular flow-based QoS (instead of bearer-based), enabling network slicing, and providing explicit support for new 5G service requirements like ultra-reliable low latency communications (URLLC) and massive Machine Type Communications (mMTC).

Without QoS, networks would operate on a simple best-effort basis, where all packets are treated equally. This is insufficient for modern digital economies that rely on real-time collaboration, industrial automation, telemedicine, and immersive entertainment. QoS solves this by allowing network operators to create service tiers, implement traffic engineering, and fulfill Service Level Agreements (SLAs). It is the technological foundation that enables the commercial promise of 5G—supporting a unified network for everything from sensors to self-driving cars.

Key Features

• Granular QoS Flow model identified by QFI, separate from the transport bearer (DRB)
• Standardized 5QI values for common service types (e.g., VoIP, video streaming, V2X messages)
• Dynamic QoS control via PCF and SMF based on application requests or network conditions
• End-to-end enforcement from UE/core to RAN with explicit packet marking (DSCP, QFI)
• Support for reflective QoS, where the UE can derive downlink QoS rules from observed traffic
• Integration with network slicing to provide isolated QoS performance per slice

Evolution Across Releases

Defining Specifications