SmartNIC & DPU Programming

High-Performance Networking & Cloud-Native · 30 illustrated lessons (text/HTML)

Outcomes

• Programme a DPU or FPGA SmartNIC data plane
• Build offloaded forwarding pipelines with DOCA Flow
• Operate SmartNICs in Kubernetes clusters

Where we teach this

• OFL-301 · SmartNIC and DPU Acceleration guided, self-paced

The lessons

Every lesson in this curriculum, in the order we teach it.

• 1Course introduction
• 2Hardware acceleration basics: from CPUs to heterogeneous computing
• 3Smart NICs and DPUs explained in depth
• 4DPU/SmartNIC hardware architecture explained
• 5Development environment setup
• 6Smart NIC programming model overview
• 7PCIe basics and device enumeration
• 8First steps with Linux kernel drivers
• 9DPDK getting started in practice
• 10SmartNIC data-plane programming with DPDK in practice
• 11Hardware offload overview
• 12Hardware offload engines
• 13Virtualisation and hardware acceleration: SR-IOV principles and practice
• 14OVS hardware offload
• 15NVMe-oF and DPU acceleration
• 16Hardware acceleration of security functions
• 17Introduction to the P4 programmable pipeline
• 18Getting started with FPGA-based SmartNIC development
• 19FPGA SmartNIC custom packet processing in practice
• 20Core performance metrics
• 21BlueField DPU in practice, part 1: getting started with the DOCA SDK
• 22BlueField DPU in practice, part 2: hardware-accelerated forwarding pipelines with the DOCA Flow API
• 23BlueField DPU in practice: accelerating AI training with DOCA and GPUDirect RDMA
• 24Exploring IPU programming
• 25Smart NICs and cloud-native: the Kubernetes device plugin mechanism in practice
• 26Deploying hardware-accelerated applications on Kubernetes in practice
• 27System-level management and orchestration: DPU firmware and lifecycle management
• 28Reliability and high availability: fault detection, isolation and recovery on DPUs/SmartNICs
• 29How major cloud providers use SmartNICs and DPUs
• 30Course summary and future outlook