Engineering Architecture Brief

NVIDIA DOCA™ (Data Center-on-a-Chip Architecture) is a comprehensive software framework and SDK designed for building high-performance, programmable applications on NVIDIA BlueField DPUs. DOCA enables developers to offload and accelerate networking, security, storage, and telemetry services from the CPU to the DPU, improving efficiency and scalability in AI factories, cloud data centers, and HPC environments. It provides APIs, libraries, drivers, and tools that expose hardware acceleration capabilities such as RDMA, packet processing, encryption, service chaining, and flow offload. DOCA is a core part of NVIDIA’s data center infrastructure stack, enabling software-defined, hardware-accelerated networking.

NV-DOCA-SDK-01

NVIDIA DOCA™ Software Framework

Programmable infrastructure software framework for accelerating networking, storage, security, and AI workloads on NVIDIA BlueField DPUs.

System Metrics Table

Acceleration Domains Networking, Storage, Security, Telemetry

AI Optimization Infrastructure Acceleration for AI Factorie

APIs C, C++, Python and REST APIs

Architecture Data Center-on-a-Chip Architecture (DOCA)

Deployment Model On-Premises, Cloud and Edge Data Centers

Developer Tools DOCA SDK, Libraries, Sample Applications and Debugging Tools

Ecosystem Integration NVIDIA BlueField, ConnectX, Spectrum, Quantum Platforms

Performance Benefit Reduced CPU Utilization and Higher Infrastructure Efficiency

Platform Type Infrastructure Software Framework / SDK

Primary Purpose DPU Application Development and Hardware Acceleration

Programming Model DPU-Offload and Service Chaining Architecture

Scalability Hyperscale Data Center Deployment Support

Security Features Encryption, Firewall Offload, Secure Boot Integration

Security Model Hardware-Accelerated Zero-Trust Infrastructure

Storage Acceleration NVMe-oF, Storage Offload Engines

Target Hardware NVIDIA BlueField DPUs

Telemetry Real-Time Infrastructure Observability

Use Cases AI Infrastructure, Cloud Networking, Cybersecurity, HPC, Data Analytics