Data Burst

Description

A Data Burst is a concept introduced in 3GPP Release 18 to characterize a specific traffic pattern where an application generates and transmits a set of multiple Protocol Data Units (PDUs) within a short time window. This definition is formally captured in TS 23.501, which specifies the system architecture for the 5G System (5GS). The concept is particularly relevant for emerging applications such as extended reality (XR), cloud gaming, and industrial automation, which often produce bursty traffic rather than a steady stream. Understanding and managing data bursts is crucial for optimizing network performance, ensuring low latency, and maintaining quality of service (QoS) for these demanding applications.

In the 5G architecture, data bursts are handled at various layers, from the application down to the physical layer. The application generates a burst of data, which is then segmented into PDUs and passed down the protocol stack. The 5G system, with its flexible numerology and QoS framework, can be configured to efficiently transport these bursts. For instance, the network can allocate resources dynamically to accommodate the burst, using techniques such as scheduling grants that match the burst size and timing. The concept is also linked to the QoS model, where specific 5G QoS Identifiers (5QIs) and QoS flows can be tailored for bursty traffic, ensuring that the burst is delivered with the required reliability and latency.

The introduction of the Data Burst concept in Rel-18 reflects the ongoing evolution of 5G to support new verticals and use cases. It is part of a broader effort to enhance the 5G system for time-sensitive communications and deterministic networking. The specifications 38.300 (NR overall description) and 38.331 (RRC protocol) provide the radio access network (RAN) aspects, detailing how the NR air interface handles bursty traffic. For example, the RAN can employ techniques like configured grants, which pre-allocate periodic resources, to reduce signaling overhead and latency for recurring bursts. Additionally, the concept informs the design of buffer status reporting and scheduling algorithms to better match the bursty nature of the traffic.

From a network management perspective, data bursts can be characterized by parameters such as burst size, burst arrival time, and burst periodicity. These parameters can be used by the network to predict and prepare for bursts, optimizing resource allocation. The 5G core network, particularly the Session Management Function (SMF) and Policy Control Function (PCF), can use this information to establish appropriate QoS flows and policies. The User Plane Function (UPF) may also employ techniques like packet buffering and forwarding to smooth out bursts or ensure timely delivery. Overall, the Data Burst concept enables a more nuanced approach to traffic handling, moving beyond simple average rate metrics to capture the temporal characteristics of modern applications.

The evolution of the Data Burst definition shows a minor change in Rel-20, where the reference to TS 23.501 was removed from the definition text. This suggests that the term has become sufficiently established that it no longer requires an explicit citation in its definition, or that the definition is now considered self-contained. The core meaning remains the same: a set of multiple PDUs generated and sent by the application in a short period of time. This stability indicates that the concept is well-understood and its fundamental characterization is unlikely to change. As 5G-Advanced and 6G evolve, the Data Burst concept will likely continue to play a role in optimizing networks for bursty, latency-sensitive applications.

Purpose & Motivation

The Data Burst concept was created to address the specific challenges posed by applications that generate traffic in bursts rather than as a continuous stream. Traditional network design often assumed relatively steady traffic patterns, using average data rates to dimension resources and set QoS parameters. However, many modern applications, such as augmented reality, virtual reality, and real-time control, produce data in short, intense bursts followed by periods of inactivity. These bursty patterns can lead to inefficient resource utilization if the network is not designed to handle them, resulting in either wasted capacity during idle periods or increased latency and packet loss during bursts.

By formally defining a Data Burst, 3GPP provides a common language and framework for both application developers and network operators to describe and manage this traffic. It allows the network to apply more sophisticated scheduling and resource allocation strategies that are aware of the bursty nature of the traffic. For example, the network can allocate resources on demand for the duration of the burst, rather than maintaining a constant allocation that would be underutilized. This leads to more efficient use of spectrum and energy, while also improving the user experience by reducing latency and jitter for bursty applications.

Historically, the need for such a concept emerged with the advent of 5G and its promise to support ultra-reliable low-latency communication (URLLC) and enhanced mobile broadband (eMBB) for new industries. Previous generations like LTE had some mechanisms to handle bursty traffic, but they were not as flexible or as explicitly defined. The Data Burst concept in Rel-18 is part of a broader set of enhancements in 5G-Advanced to support time-sensitive networking (TSN), deterministic networking, and XR. It builds on earlier work on QoS and traffic models, providing a more precise tool for optimizing network performance for the most demanding applications.