Grafana
Telemetry & ML Defense Monitoring
The visualization layer for our threat intelligence pipeline. Tracking model inference latency, honeypot attack vectors, and autonomous defense actions in real-time.
Validating model performance.
Our deployment of Grafana goes beyond standard system administration; it is the primary lens through which we analyze the performance of local LLMs and the efficacy of our ML-driven threat detection pipelines. By correlating inference latency with GPU power states, we optimize architectures for edge deployment.
Key References
Turnbull, J. (2018). "Monitoring with Prometheus." Turnbull Press.
Prometheus → Grafana pipeline.
Node exporters and cAdvisor collect metrics from both Hub and Satellite nodes. Prometheus scrapes every 15 seconds, Grafana visualizes with custom dashboards, and Uptime Kuma monitors external availability from the edge.
Everything. Monitored.
Docker Overview
Container health, resource usage, restart counts across 42+ services.
System Resources
CPU, RAM, disk I/O, network throughput for Hub and Satellite nodes.
Security Events
Suricata alerts, CrowdSec bans, honeypot activity, Wazuh SIEM events.
GPU & AI Metrics
VRAM usage, GPU temperature, model inference latency, Ollama throughput.
Service Health
Uptime tracking, response times, SSL cert expiry, DNS resolution.
Alert Rules
Active alerts, firing history, notification routing to ntfy channels.