nomad-hpc 1.2.6

A lightweight HPC monitoring and predictive analytics tool

NØMAD-HPC

NØde Monitoring And Diagnostics — Lightweight HPC monitoring, visualization, and predictive analytics.

"Travels light, adapts to its environment, and doesn't need permanent infrastructure."

---

📖 Full Documentation — Installation guides, configuration, CLI reference, network methodology, ML framework, and more.

---

Quick Start

pip install nomad-hpc
nomad demo                    # Try with synthetic data

For production:

nomad init                    # Configure for your cluster
nomad collect                 # Start data collection
nomad dashboard               # Launch web interface

---

Features

Feature	Description	Command
Dashboard	Real-time multi-cluster monitoring with partition views	nomad dashboard
Workstation Monitoring	Track departmental workstations (CPU, memory, disk, users)	Dashboard → Workstations
Storage Monitoring	Monitor NFS servers, ZFS pools, IOPS, and client connections	Dashboard → Storage
Interactive Sessions	Monitor RStudio/Jupyter sessions with memory and age	Dashboard → Interactive
Data Readiness	Assess ML model readiness with sample size and variance analysis	nomad readiness
Diagnostics	Analyze network, storage, and node-level bottlenecks	nomad diag
Educational Analytics	Track computational proficiency development	nomad edu explain <job>
Alerts	Threshold + predictive alerts (email, Slack, webhook)	nomad alerts
ML Prediction	Job failure prediction using similarity networks	nomad predict
Community Export	Anonymized datasets for cross-institutional research	nomad community export

---

Architecture

┌─────────────────────────────────────────────────────────────────────┐
│                              NØMAD                                  │
├───────────────┬───────────────┬───────────────┬─────────────────────┤
│  Collectors   │   Analysis    │     Viz       │      Alerts         │
├───────────────┼───────────────┼───────────────┼─────────────────────┤
│ disk          │ derivatives   │ dashboard     │ thresholds          │
│ iostat        │ similarity    │ network 3D    │ predictive          │
│ slurm         │ ML ensemble   │ partitions    │ email/slack         │
│ gpu           │ edu scoring   │ workstations  │ webhooks            │
│ nfs           │ readiness     │ storage       │                     │
│ workstation   │ diagnostics   │ interactive   │                     │
│ storage       │               │               │                     │
└───────────────┴───────────────┴───────────────┴─────────────────────┘
                                │
                      ┌─────────┴─────────┐
                      │  SQLite Database  │
                      └───────────────────┘

---

CLI Reference

Core Commands

nomad init                    # Setup wizard
nomad collect                 # Start collectors
nomad dashboard               # Web interface
nomad dashboard --db file.db  # Use specific database
nomad demo                    # Demo mode with synthetic data
nomad status                  # System status

Data Readiness & Diagnostics

nomad readiness               # Check ML training readiness
nomad readiness -v            # Verbose with feature details
nomad diag network            # Network performance analysis
nomad diag storage            # Storage health and I/O patterns
nomad diag node               # Node-level resource bottlenecks

Educational Analytics

nomad edu explain <job_id>    # Job analysis with recommendations
nomad edu trajectory <user>   # User proficiency over time
nomad edu report <group>      # Course/group report

Analysis & Prediction

nomad disk /path              # Filesystem trends
nomad jobs --user <user>      # Job history
nomad similarity              # Network analysis
nomad train                   # Train ML models
nomad predict                 # Run predictions

Community & Alerts

nomad community export        # Export anonymized data
nomad community preview       # Preview export
nomad alerts                  # View alerts
nomad alerts --unresolved     # Unresolved only

---

Dashboard Views

The web dashboard includes multiple views accessible via tabs:

• Cluster Overview: Real-time node status with health rings showing CPU utilization
• Network View: 3D job similarity network with failure clustering analysis
• Resources: CPU-hours, GPU-hours, and usage breakdown by group/user
• Activity: Job submission heatmap showing patterns by day and hour
• Interactive: Active RStudio and Jupyter sessions with memory usage
• Workstations: Departmental machines with CPU, memory, disk, and logged-in users
• Storage: NFS servers with ZFS pool health, capacity, and client connections

Toggle between light and dark themes with the Theme button.

---

Installation

From PyPI

pip install nomad-hpc

From Source

git clone https://github.com/jtonini/nomad-hpc
cd nomad-hpc && pip install -e .

Requirements

• Python 3.9+
• SQLite 3.35+
• sysstat package (iostat, mpstat)
• Optional: SLURM, nvidia-smi, nfsiostat

System Check

nomad syscheck

---

Documentation

📖 jtonini.github.io/nomad-hpc

• Installation & Configuration
• System Install (--system)
• Dashboard Guide
• Educational Analytics
• Network Methodology
• ML Framework
• Proficiency Scoring
• CLI Reference
• Configuration Options

---

License

Dual-licensed:

• AGPL v3 — Free for academic, educational, and open-source use
• Commercial License — Available for proprietary deployments

---

Citation

@software{nomad2026,
  author = {Tonini, João Filipe Riva},
  title = {NØMAD: Lightweight HPC Monitoring with Machine Learning-Based Failure Prediction},
  year = {2026},
  url = {https://github.com/jtonini/nomad-hpc},
  doi = {10.5281/zenodo.18614517}
}

---

Contributing

See CONTRIBUTING.md for guidelines.

---

Contact

• Author: João Tonini
• Email: [e-mail]
• Issues: GitHub Issues