Network Based Media Processing

Description

Network Based Media Processing (NBMP) is a comprehensive framework standardized by 3GPP, starting in Release 16, that enables the deployment, orchestration, and lifecycle management of media processing workloads within a telecommunications network. It treats media processing functions—such as video transcoding, audio mixing, object detection, spatial audio rendering, and quality analysis—as modular, reusable, and network-accessible services. The core idea is to move media processing from rigid, dedicated appliances at the edge of the network or in end-user devices to a flexible, cloud-native environment that can be dynamically instantiated and chained to create media workflows.

The NBMP architecture is built around several key components defined in the 3GPP specifications. The central entity is the NBMP Workflow Manager, which is responsible for creating, configuring, and managing media processing workflows. A workflow is described using a Media Processing Description (MPD), which is an XML or JSON document that defines the source of the media, the sequence of media processing functions (called Media Functions or MFs) to be applied, their configuration parameters, and the destination for the processed output. The Workflow Manager interacts with a Media Function Repository, which stores descriptors for available MFs, and a Media Function Instance Manager, which is responsible for instantiating the actual software containers or virtual machines that execute the MFs on compute resources (e.g., in a MEC host or cloud data center).

NBMP works by allowing a service provider or application developer to submit a workflow description to the Workflow Manager via a northbound API. The Workflow Manager then parses this description, resolves the required Media Functions from the repository, and instructs the Media Function Instance Manager to deploy them on suitable compute resources with the necessary networking connectivity. It establishes the data flows between the source, the chain of MFs, and the sink. The framework also includes capabilities for monitoring the status and performance of the workflow, applying updates, and scaling instances up or down based on load. This allows for the creation of complex, real-time media pipelines that can adapt to network conditions, user device capabilities, and application requirements, which is crucial for delivering high-quality, interactive media experiences over 5G.

Purpose & Motivation

NBMP was created to address the growing complexity and demand for advanced media services in the 5G era. Traditional media delivery relied on pre-processed content or heavy processing in end-user devices, which limited flexibility, scalability, and the ability to introduce new features. The rise of applications like cloud gaming, volumetric video for AR/VR, real-time video analytics, and personalized broadcast (e.g., different bitrates, resolutions, overlays for different viewers) required a new paradigm where media could be dynamically processed within the network itself.

The primary problems NBMP solves are vendor lock-in, operational rigidity, and inefficient resource utilization. Before NBMP, deploying a new media service often required integrating proprietary hardware and software from a specific vendor, making it difficult to innovate or scale. NBMP standardizes the interfaces and descriptions for media functions, promoting interoperability and a multi-vendor ecosystem. It also solves the problem of latency and bandwidth for immersive media; by processing media closer to the user at the network edge (e.g., in a Multi-access Edge Computing node), NBMP can reduce latency for interactive applications and save core network bandwidth by adapting a single high-quality stream into multiple optimized versions near the consumption point.

Furthermore, NBMP is a key enabler for network slicing and service automation. A media processing workflow can be instantiated as part of a network slice dedicated to, for example, a live event broadcast or an esports tournament. The framework's ability to dynamically manage the lifecycle of these workflows allows operators to offer Media-as-a-Service, creating new revenue streams and meeting the stringent quality-of-service requirements of next-generation media applications in a cost-effective and agile manner.

Key Features

• Standardized Media Processing Description (MPD) for declarative workflow definition
• Modular and reusable Media Functions (MFs) with standardized descriptors
• Dynamic orchestration and lifecycle management of media processing workloads
• Support for deployment on heterogeneous compute platforms (cloud, MEC)
• Integration with 5G network capabilities for QoS, network slicing, and edge computing
• Northbound APIs for service providers to create and manage media workflows

Evolution Across Releases

Defining Specifications