Service Model

Description

The Service Model is a foundational concept in 3GPP QoS architecture, defined in TS 21.905 (Vocabulary) and elaborated in TS 22.060 (GPRS Service Description). It provides a general characterisation of services based upon a QoS paradigm, deliberately without specifying the actual performance targets. This abstraction allows the same service model to be applied across different access technologies, releases, and deployment scenarios while leaving the concrete parameter values to be negotiated at session setup or configured by the operator.

The service model operates by classifying traffic into service classes or traffic classes, each associated with a set of QoS attributes such as traffic handling priority, allocation/retention priority, guaranteed bit rate, maximum bit rate, transfer delay, and error ratio. In the GPRS and UMTS era, TS 22.060 defined four traffic classes: Conversational, Streaming, Interactive, and Background. Each class has distinct QoS characteristics — for example, Conversational is delay-sensitive and symmetric, while Background is delay-tolerant and used for best-effort applications. The service model does not itself fix numeric values; instead, it defines the qualitative expectations that the network must meet.

Architecturally, the service model sits between the application layer and the bearer service layer. Applications or UEs request a service class, and the network maps that request onto bearer capabilities. In EPS (LTE), this evolved into the QCI (QoS Class Identifier) framework defined in TS 23.203, where each QCI maps to a service model with specific packet delay budgets, packet error loss rates, and resource type (GBR or non-GBR). In 5G, the 5QI (5G QoS Identifier) extends this further, adding reflective QoS, notification control, and flow-based QoS. Despite these evolutions, the underlying principle remains: the service model characterises the service qualitatively, and the concrete parameters are applied by the bearer or QoS flow.

The service model is essential for interoperability. Because it is defined at a high level of abstraction, it allows different vendors, operators, and access technologies to implement QoS consistently. It also enables roaming, since the visited network can interpret the service model without needing proprietary details from the home network. The model supports negotiation: the UE requests a service class, the network may accept, modify, or reject the request based on available resources and policy.

In practice, the service model is realised through bearer services in UMTS/GPRS, EPS bearers in LTE, and QoS flows in 5G. Each of these maps the abstract service model onto concrete transport mechanisms. The service model thus serves as the contract between the application's QoS expectations and the network's transport capabilities, ensuring that services are delivered with appropriate quality without over-specifying implementation details.

Purpose & Motivation

The Service Model exists to provide a technology-agnostic, qualitative framework for describing the QoS requirements of services. Before its introduction, QoS was often described in terms of specific numeric parameters tied to particular access technologies, which made interoperability and roaming difficult. The service model solves this by separating the qualitative characterisation of a service from the quantitative parameters used to implement it.

Historically, early packet-switched networks such as GPRS needed a way to differentiate traffic types — voice, video, web browsing, and background downloads — without hard-coding values that would become obsolete as technology evolved. The service model, introduced in Rel-4 and documented in TS 22.060, provided this abstraction. It allowed operators to define service classes that could be mapped to different bearer types depending on the radio access technology (GERAN, UTRAN, later E-UTRAN and NR).

The model also addresses the problem of negotiation and policy control. By defining services qualitatively, the network can apply admission control, resource reservation, and policy decisions without being locked into specific parameter values. This flexibility is crucial for supporting diverse applications and for enabling the same service model to be reused across multiple generations of technology, from GPRS through UMTS, LTE, and 5G.