DOCA

DOCA, short for NVIDIA DOCA, is a software development framework for data processing units (DPUs) used in modern data centers. It enables developers to build and run offloaded workloads on the DPU, including networking, storage, and security tasks, thereby reducing the load on the host CPU and increasing overall performance and efficiency.

The DOCA platform provides libraries, application programming interfaces, and runtime components that expose the DPU’s resources through higher-level abstractions. Core elements include a DOCA SDK with sample applications, DOCA libraries for common offload patterns, and a DOCA Manager to coordinate resource allocation and lifecycle management. The framework is designed to work with Linux on the host and supports containerized deployment for modern orchestration.

Hardware and ecosystem: DOCA is tightly integrated with NVIDIA BlueField DPUs, which combine processing cores, programmable NIC capabilities, and cryptographic accelerators. DOCA programs run on the DPU’s hardware and interact with host software via standardized interfaces, enabling features such as packet processing, encryption, and policy enforcement to be offloaded from the host.

Use cases: Network function virtualization and software-defined networking, firewall and VPN services, TLS/crypto offload, storage I/O offload, virtualized networking and telemetry, and other data-plane acceleration tasks in data centers and cloud environments.

History and development: NVIDIA introduced DOCA as part of its DPU software ecosystem in the 2020s to provide a unified development environment for DPU-based acceleration and security workloads. It is designed to streamline the creation and deployment of DPU-accelerated services.