jerome
b671546622
Created complete documentation structure with: • Root README.md with repository overview • Documentation directory structure (01-05 + templates) • Migrated content from public repository • New operational documentation • Comprehensive templates for business workflows Sections: 1. 01-business-strategy/ - Business vision, market analysis, financials 2. 02-technical-architecture/ - Infrastructure, AI agents, deployment 3. 03-operations/ - Client onboarding, support processes 4. 04-customer-portal/ - Placeholder for client-facing docs 5. 05-development/ - Placeholder for dev standards 6. templates/ - 5 reusable business templates This commit establishes the foundation for the repository reorganization. Content review will be conducted separately.
1.5 KiB
1.5 KiB
1|# 02 - GPU Monitoring — Home
2|**Purpose:** Grafana dashboards and monitoring setup for GPU telemetry.
3|
4|## Contents
5|
6|This directory contains:
7|- Grafana dashboard configuration
8|- ROCm monitoring setup
9|- Real-time performance tracking
10|
11|### Monitoring Features
12|
13|- **Real-time GPU utilization:** Track VRAM, compute, and memory usage
14|- **Temperature monitoring:** Alert on overheating conditions
15|- **Performance metrics:** Core clock, power consumption, fan speeds
16|- **Historical data:** Grafana time-series graphs for trend analysis
17|
18|## Quick Start
19|
20|### Grafana Dashboard Setup
21|
22|```bash
23|# Install Grafana
24|sudo apt install grafana
25|
26|# Start Grafana service
27|systemctl start grafana-server
28|
29|# Access dashboard
30|# http://localhost:3000 (admin/admin)
31|```
32|
33|### ROCm Monitoring
34|
35|```bash
36|# Install ROCm health monitoring tools
37|sudo apt install rocminfo rocm-smi
38|
39|# Real-time monitoring
40|watch -n 1 rocm-smi
41|```
42|
43|## Related Infrastructure
44|
45|- **[AMD GPU Passthrough](../03\ -\ AMD\ GPU\ Passthrough/)** — Hardware configuration
46|- **[Qwen3.5 Model Setup](../04\ -\ Qwen3.5\ Model\ Setup/)** — Model deployment
47|
48|---
49|
50|*Monitor your GPUs 24/7 to ensure optimal performance for LLM inference workloads.*
51|