Intel Tiber Cloud Deployment
Deploy TrustGraph on Intel Tiber Cloud with Intel GPU and Gaudi accelerated systems for high-performance AI workloads.
Overview
TrustGraph provides specialized deployment configurations for Intel’s advanced hardware platforms, including Intel Tiber Cloud and bare-metal Intel GPU/Gaudi systems. This deployment method is designed for:
- High-performance AI inference with Intel accelerators
- Self-hosted deployments on Intel hardware
- Research and development with cutting-edge Intel AI technologies
- Enterprise workloads requiring Intel-optimized performance
⚠️ Work in Progress: This deployment method is actively being developed and optimized for Intel’s latest hardware platforms.
What You Get
The Intel Tiber Cloud deployment includes:
- Intel accelerated systems (Gaudi, GPU, or GR platforms)
- Optimized AI inference with vLLM or TGI servers
- Large language models like Llama 3.3 70B
- Complete TrustGraph stack deployed via containerization
- SSH-based deployment with automated scripts
- Port forwarding setup for secure access
- Monitoring and observability with Grafana
- Web workbench for document processing and Graph RAG
Intel Hardware Platforms
Intel Gaudi Systems
- Software: TrustGraph 1.0.13 + vLLM (HabanaAI fork)
- Model: Llama 3.3 70B
- Deployment:
deploy-gaudi-vllm.zip
- Optimized for: AI training and inference workloads
Intel GPU Multi-GPU 1550 Systems
- Software: TrustGraph 1.0.13 + TGI Server 3.3.1-intel-xpu
- Deployment:
deploy-gpu-tgi.zip
- Optimized for: High-throughput GPU inference
Intel GR Systems
- Software: TrustGraph 1.0.13 + TGI Server 3.3.1-intel-xpu
- Deployment:
deploy-gr.zip
- Optimized for: Specialized Intel AI workloads
Why Choose Intel Tiber Cloud?
🚀 Cutting-Edge AI Hardware
- Intel Gaudi: Purpose-built for AI training and inference
- Intel GPU: High-performance parallel processing
- Specialized Architecture: Optimized for AI/ML workloads
⚡ Performance Optimization
- Hardware-Accelerated Inference: Native Intel optimization
- Large Model Support: Handle models like Llama 3.3 70B efficiently
- Reduced Latency: Direct hardware acceleration
🔒 Self-Hosted Control
- Data Sovereignty: Complete control over data and models
- Custom Configuration: Tailor deployments to specific needs
- Enterprise Security: Self-hosted infrastructure
🛠️ Developer Access
- Research Platform: Access to latest Intel AI technologies
- Experimentation: Test advanced AI configurations
- Direct Hardware Access: Low-level optimization capabilities
Deployment Method
The Intel deployment uses automated deployment scripts that:
- Connect via SSH jump host to Intel Tiber Cloud
- Deploy pre-configured TrustGraph packages
- Set up Intel-optimized AI inference servers
- Configure port forwarding for secure access
- Initialize monitoring and web interfaces
Quick Process Overview
- Obtain access to Intel Tiber Cloud instance
- Create HuggingFace token and accept model licenses
- Choose deployment type (Gaudi, GPU, or GR)
- Deploy via script with SSH parameters
- Connect via SSH with port forwarding
- Access services through forwarded ports
Access Configuration
Intel Tiber Cloud uses SSH jump host access:
# SSH connection format
ssh -J guest@[jump-host] sdp@[target-host]
# Deployment command
./deploy-tiber guest@[jump-host] sdp@[target-host] [deployment-package]
# Port forwarding
./port-forward guest@[jump-host] sdp@[target-host]
Access Points
Once deployed, you’ll have access to:
- TrustGraph API: Port 8089 (forwarded from 8088)
- Web Workbench: Port 8889 (forwarded from 8888)
- Grafana Monitoring: Port 3001 (forwarded from 3000)
Model Support
Large Language Models: Llama 3.3 70B and other HuggingFace models License Requirements: HuggingFace account with model access Hardware Optimization: Intel-specific optimizations for inference Inference Engines: vLLM (Gaudi) and TGI (GPU/GR)
Complete Documentation
For detailed step-by-step instructions, deployment packages, and troubleshooting, visit:
TrustGraph Intel Tiber Cloud Guide
The repository contains:
- Deployment automation scripts
- Intel platform-specific configurations
- Pre-built deployment packages
- SSH connection utilities
- Port forwarding setup
- Detailed setup instructions
- Intel hardware optimization guides
Prerequisites
Intel Tiber Cloud Access: Account and instance allocation HuggingFace Token: For model downloads and licensing SSH Access: Familiarity with SSH jump host connections Model Licenses: Acceptance of required model licenses (e.g., Llama)
Performance Benefits
Hardware Acceleration: Native Intel AI acceleration Large Model Efficiency: Optimized for 70B+ parameter models Reduced Infrastructure Costs: Efficient hardware utilization Custom Optimization: Direct access to hardware-level tuning
Use Cases
Research & Development: Access to cutting-edge Intel AI hardware High-Performance Inference: Large model deployment with optimal performance Self-Hosted AI: Complete control over AI infrastructure Intel Ecosystem Integration: Leverage Intel’s AI software stack
Next Steps
After deployment, you can:
- Load documents through the web workbench
- Test Graph RAG queries with Llama 3.3 70B
- Monitor Intel hardware performance through Grafana
- Experiment with Intel-optimized AI configurations
- Benchmark performance against other platforms
- Integrate with Intel’s AI development tools