Application User Guide

• 1. OVS Offload
  • 1.1. Introduction
  • 1.2. Architecture
    • 1.2.1. Packet Flow - Wire-VM Communication and vice versa
  • 1.3. Getting the sources and compile
  • 1.4. Setting up the Environment
    • 1.4.1. Binding required devices
  • 1.5. Launching the application
    • 1.5.1. Understanding the output
    • 1.5.2. Understanding different nodes
  • 1.6. Various Packet Flow Scenarios (with demo)
    • 1.6.1. Wire-VM Communication
• 2. DAO Crypto Agent App
  • 2.1. Overview
  • 2.2. Architecture Summary
  • 2.3. Lifecycle
  • 2.4. Host Interaction
  • 2.5. Validation
  • 2.6. Build & Run
  • 2.7. References
• 3. VirtIO-crypto
  • 3.1. Setting up EP environment
    • 3.1.1. Setup SDP PF/VF count in EBF menu
    • 3.1.2. Setup huge pages for DPDK application
    • 3.1.3. Bind required DMA devices to vfio-pci
    • 3.1.4. Bind required CPT VF’s to vfio-pci
    • 3.1.5. Bind PEM BAR4 and DPI BAR0 platform devices to vfio-platform
    • 3.1.6. Bind required SDP RVU devices to vfio-pci
  • 3.2. Running the EP firmware application
    • 3.2.1. EAL Options
    • 3.2.2. Application Options
    • 3.2.3. Example EP firmware command
  • 3.3. Setting up Host environment
  • 3.4. Running DPDK crypto-perf on host virtio device
• 4. VirtIO-l2fwd
  • 4.1. Setting up EP environment
    • 4.1.1. Setup SDP PF/VF count in EBF menu
    • 4.1.2. Setup huge pages for DPDK application
    • 4.1.3. Bind required DMA devices to vfio-pci
    • 4.1.4. Bind required RPM VF’s to vfio-pci
    • 4.1.5. Bind PEM BAR4 and DPI BAR0 platform devices to vfio-platform
    • 4.1.6. Bind required SDP RVU devices to vfio-pci
  • 4.2. Running the EP firmware application
    • 4.2.1. EAL Options
    • 4.2.2. Application Options
    • 4.2.3. Example EP firmware command
  • 4.3. Setting up Host environment
• 5. VirtIO-blkIO
  • 5.1. Architecture
  • 5.2. Application Workflow
    • 5.2.1. Initialization
    • 5.2.2. Main Processing Loop
    • 5.2.3. Request Processing
    • 5.2.4. Completion Handling
    • 5.2.5. Teardown
  • 5.3. Setting up EP environment
    • 5.3.1. Setup SDP PF/VF count in EBF menu
    • 5.3.2. Setup huge pages for DPDK application
    • 5.3.3. Bind required DMA devices to vfio-pci
    • 5.3.4. Bind required NPA PF to vfio-pci
    • 5.3.5. Bind required SDP RVU devices to vfio-pci
    • 5.3.6. Running the EP firmware application
  • 5.4. Setting up Host environment
  • 5.5. Using block device on host
• 6. Secgw-graph
  • 6.1. Features
  • 6.2. Setting up environment
    • 6.2.1. Setup of SDP and RPM devices
    • 6.2.2. Setup huge pages for DPDK application
    • 6.2.3. Bind required RPM VF’s to vfio-pci
    • 6.2.4. Bind required SDP VF’s to vfio-pci
    • 6.2.5. Bind required CPT VF to vfio-pci
    • 6.2.6. Bind required Inline device PF to vfio-pci
    • 6.2.7. Setting up Host environment
  • 6.3. Running secgw-graph
    • 6.3.1. EAL Options
    • 6.3.2. Application Options
    • 6.3.3. secgw.cli
    • 6.3.4. Application bootup logs
    • 6.3.5. CLI terminal
  • 6.4. CLI commands
  • 6.5. Running as IPv4 router
    • 6.5.1. Configuration
    • 6.5.2. ping host2 IP from host1
    • 6.5.3. How it works
    • 6.5.4. Resolving Partial Routes
  • 6.6. Running as IPsec gateway
    • 6.6.1. Configuration
    • 6.6.2. Check IPsec configuration
    • 6.6.3. ping host2 IP from host1
  • 6.7. Planned Features
  • 6.8. Known Issues
• 7. Universal SmartNIC Turnkey App
  • 7.1. Introduction
    • 7.1.1. Feature List
  • 7.2. Getting the sources and compile
    • 7.2.1. Binding required devices
  • 7.3. Launching the application
    • 7.3.1. Application Options
    • 7.3.2. Example Commands
    • 7.3.3. Supported CLI commands
    • 7.3.4. Runtime configuration
    • 7.3.5. Application running demo
• 8. FD.io Vector Packet Processing (VPP)
  • 8.1. What is VPP?
  • 8.2. VPP Accelerations
  • 8.3. Marvell’s High-Performance VPP for OCTEON
    • 8.3.1. Installing and Running VPP for Packet I/O Acceleration Demo
    • 8.3.2. Running VPP for Crypto offload Demo
    • 8.3.3. VPP FD.io Documentation
  • 8.4. VPP Usage and Integrations
  • 8.5. Compiling VPP from sources
  • 8.6. Enhancing VPP IPsec offload with Strongswan Integration
    • 8.6.1. Strongswan Integration
    • 8.6.2. Configuring VPP IPsec route mode or policy mode
    • 8.6.3. VPP startup.conf configuration and CLI commands
    • 8.6.4. Strongswan configuration on DPU
    • 8.6.5. Strongswan configuration on remote DPU/Host:
    • 8.6.6. Ipsec Linux commands
  • 8.7. Host Offload to OCTEON DPU Via Virtio Interface over PCIe
    • 8.7.1. Configuring OCTEON DPU
  • 8.8. Host offload to OCTEON DPU via SDP interface over PCIe
    • 8.8.1. Configuring OCTEON DPU
    • 8.8.2. Configuring Host
• 9. TLS Proxy with NGINX
  • 9.1. Introduction
  • 9.2. NGINX Brief
  • 9.3. NGINX as TLS termination proxy
    • 9.3.1. Usage/deployment Model
    • 9.3.2. NGINX Architecture Diagram
    • 9.3.3. NGINX acceleration on OCTEON
  • 9.4. Steps to Install
    • 9.4.1. Ubuntu Debian package
    • 9.4.2. Compiling from Sources
  • 9.5. Environment Setup
    • 9.5.1. Run the below script (once after every reboot) to create CPT VFs and bind to vfio-pci:
    • 9.5.2. NGINX configuration
    • 9.5.3. OpenSSL configuration
  • 9.6. Launching the application
    • 9.6.1. Running the Proxy Application
    • 9.6.2. Functional Testing of the Proxy
    • 9.6.3. Performance Testing of the Proxy
• 10. Machine Learning
  • 10.1. Introduction
  • 10.2. Setting Up The TVM Compiler Framework
    • 10.2.1. Prerequisites
    • 10.2.2. Cloning and Building TVM Sources
    • 10.2.3. Setting MMLC Binaries
  • 10.3. Model Compilation
    • 10.3.1. Model Layers and Operators
    • 10.3.2. Compiler Options
    • 10.3.3. FP16 Compilation Flow
    • 10.3.4. Compilation Flow for CPU Execution
    • 10.3.5. Pre-processing and Post-processing Steps
  • 10.4. Running Inference
    • 10.4.1. Simulator Mode
    • 10.4.2. DPDK ML Test Application Mode
    • 10.4.3. Output Validation
  • 10.5. Example Models/Usecases
    • 10.5.1. Resnet50-v1-12 Model Compilation using Jupyter Notebook
    • 10.5.2. LUCID
    • 10.5.3. Mnist-12
    • 10.5.4. GoogLeNet
    • 10.5.5. MobileNetv2-12
    • 10.5.6. VGG11
    • 10.5.7. Resnet101_v1b
    • 10.5.8. Darknet-53
    • 10.5.9. Encrypted Traffic Classification Model
    • 10.5.10. References
• 11. Snort with VectorScan
  • 11.1. What is Snort?
  • 11.2. What is VectorScan?
  • 11.3. Accelerations
  • 11.4. Installing Snort
    • 11.4.1. Ubuntu Debian package
    • 11.4.2. Install the Snort package and verify
    • 11.4.3. Path layout
  • 11.5. VectorScan Benchmarking
    • 11.5.1. Pattern Sets
    • 11.5.2. Performance measurements
  • 11.6. Snort Benchmarking
    • 11.6.1. Configure library path
    • 11.6.2. Running Snort
• 12. Kubernetes CNI Offload
  • 12.1. Introduction
  • 12.2. Architecture of CNI-Offload Solution
    • 12.2.1. Host Components
    • 12.2.2. DPU Components
    • 12.2.3. Solution Software and Firmware Matrix
  • 12.3. Installation Steps
    • 12.3.1. System Requirements:
    • 12.3.2. Add nodes to k8s Cluster
    • 12.3.3. Build and Deploy CNI-Offload Solution
    • 12.3.4. Verify CNI-Offload Solution is Deployed
  • 12.4. Test Star Wars Demo with CNI-Offload
    • 12.4.1. Testbed Diagram:
    • 12.4.2. Deploy Services
    • 12.4.3. Service Offload Validation
    • 12.4.4. Policy Offload Validation
    • 12.4.5. Star Wars Demo Video with CNI-Offload
  • 12.5. Test IPsec Transparent Encryption with CNI-Offload
    • 12.5.1. IPSec Demo Video with CNI-Offload
• 13. Envoy Acceleration on OCTEON
  • 13.1. Overview
  • 13.2. Architecture
  • 13.3. Accelerating Envoy on Marvell OCTEON Platform
  • 13.4. Accelerating TLS handshake
  • 13.5. Private Key Provider Extension
  • 13.6. DPDK Crypto API Integration
  • 13.7. Benefits
  • 13.8. Deployment
• 14. RDMA Solution
  • 14.1. RDMA Overview
  • 14.2. RDMA Operation Modes
    • 14.2.1. Octeon RDMA End Point Mode
    • 14.2.2. Octeon RDMA Termination Mode
  • 14.3. Testing
    • 14.3.1. UD Ping-Pong Test
    • 14.3.2. Multi-Queue UD Test (ibv_rdma_mq_trf)
    • 14.3.3. Multi-Device RDMA Steps
    • 14.3.4. Troubleshooting
    • 14.3.5. Command-Line Scenarios (Multi-QP Application)
    • 14.3.6. Performance Testing (Perftest Suite)
    • 14.3.7. Latency Tests (UD / RC):
    • 14.3.8. Bandwidth Tests (UD / RC SEND, WRITE, READ):
  • 14.4. RDMA Counters
    • 14.4.1. Connection & Access Overview
    • 14.4.2. RDMA Core & Port Overview
    • 14.4.3. RDMA Port-Level Statistics
    • 14.4.4. RDMA Port & QP Overview
    • 14.4.5. Checking Per-QP Status
    • 14.4.6. RDMA Queue Pair Statistics
    • 14.4.7. RDMA Ethdev Statistics
    • 14.4.8. Counter Lifecycle
    • 14.4.9. Debug-Only Counters
  • 14.5. Planned Enhancements
  • 14.6. Known Issues
  • 14.7. References
• 15. NVMe-oF Storage Target on OCTEON
  • 15.1. Introduction
  • 15.2. Solution Components
  • 15.3. Architecture
  • 15.4. Prerequisites
    • 15.4.1. Hardware
    • 15.4.2. Software
    • 15.4.3. System configuration
  • 15.5. Building the Components
    • 15.5.1. DAO data plane
    • 15.5.2. Building the OCTEON RDMA provider on OCTEON
    • 15.5.3. Building the NVMe-oF SPDK target on OCTEON
    • 15.5.4. Building SPDK initiator tools on NVMe-oF Initiator System
    • 15.5.5. Deploying the artifacts
  • 15.6. Environment Setup
  • 15.7. Setting up the OCTEON Target
    • 15.7.1. Step 1 – launch the RDMA data plane
    • 15.7.2. Step 2 – configure the data interface and start the target
    • 15.7.3. Step 3 – configure the target over RPC
  • 15.8. Setting up the NVMe-oF Initiator
  • 15.9. Functional Testing
  • 15.10. Performance Testing
    • 15.10.1. Baseline (single core, single queue pair)
    • 15.10.2. Multi-core, multi-queue-pair testing
  • 15.11. Limitations
    • 15.11.1. Shared Receive Queue (SRQ) not supported
    • 15.11.2. Maximum NVMe queue depth of 16
    • 15.11.3. Completion queue resize (cv_resize_q) not supported
  • 15.12. Troubleshooting
    • 15.12.1. Bring-up order
    • 15.12.2. dao-rdma_graph fails or port down
    • 15.12.3. octep_rdma_0 not found
    • 15.12.4. nvmf_tgt / RPC fails
    • 15.12.5. Initiator cannot connect