Educational Analytics¶

NØMAÐ Edu bridges the gap between infrastructure monitoring and educational outcomes, helping instructors, mentors, and users track the development of computational proficiency.

Overview¶

Traditional HPC metrics tell you what happened. NØMAÐ Edu tells you how well users are learning to use HPC effectively.

Use cases:

Instructors: Track class-wide skill development, identify struggling students
Research mentors: Monitor graduate student onboarding progress
HPC staff: Evaluate workshop and training effectiveness
Users: Self-assess and improve HPC practices

Quick Start¶

# Explain a job with proficiency scores and recommendations
nomad edu explain 12345

# Track a user's improvement over time
nomad edu trajectory alice

# Generate a report for a course or research group
nomad edu report cs301

# Your own jobs: score, trend and what to change
nomad edu me

Commands¶

`nomad edu explain`¶

Analyze a single job with proficiency scores and actionable recommendations.

nomad edu explain <job_id> [options]

Options:

Option	Description
`--db PATH`	Database path (default: configured DB)
`--json`	Output as JSON
`--no-progress`	Skip historical comparison

Example output:

  NØMAÐ Job Analysis — 1104
  ────────────────────────────────────────────────────────
  User: alice    Partition: compute    Node: node03
  State: COMPLETED    Runtime: 33h 38m / 48h 00m requested

  Proficiency Scores
  ────────────────────────────────────────────────────────
    CPU Efficiency       ██░░░░░░░░   22.6%   Needs Work
    Memory Efficiency    █████████░   90.4%   Excellent
    Time Estimation      ██████████   97.4%   Excellent
    I/O Awareness        ███████░░░   68.8%   Good
    ────────────────────────────────────────────────────
    Overall Score        ███████░░░   69.8%   Good

  Recommendations
  ────────────────────────────────────────────────────────
    CPU Efficiency:
      Very low CPU utilization at 21% — requested 4
      cores but used ~1. This wastes resources and
      may delay other users' jobs.
      Try: #SBATCH --ntasks=1
          If your code is single-threaded, request 1 core.

  Your Progress (last 30 jobs)
  ────────────────────────────────────────────────────────
    CPU Efficiency        53.7% →  22.6%  ↓ declining
    Memory Efficiency     90.0% →  90.4%  → stable
    Time Estimation       88.4% →  97.4%  ↑ improving
    I/O Awareness         91.8% →  68.8%  ↓ declining

`nomad edu trajectory`¶

Track a user's proficiency development over time.

nomad edu trajectory <username> [options]

Options:

Option	Description
`--db PATH`	Database path
`--days N`	Lookback period (default: 90)
`--json`	Output as JSON

Example output:

  NØMAÐ Proficiency Trajectory — alice
  ────────────────────────────────────────────────────────
  Jobs analyzed: 173    Period: 2026-02-04 → 2026-02-15
  Stable proficiency

  Score Progression
  ────────────────────────────────────────────────────────
    2026-01-29    ████████░░   78.6%  (21 jobs)
    2026-02-05    █████████░   78.9%  (144 jobs)

  Dimension Changes
  ────────────────────────────────────────────────────────
    I/O Awareness         90.6%  → +4.6%
    Memory Efficiency     85.9%  → +0.2%
    GPU Utilization       85.0%  → +0.0%
    CPU Efficiency        51.9%  → -1.2%
    Time Estimation       81.3%  → -2.1%

`nomad edu report`¶

Generate aggregate reports for courses, research groups, or any Linux group.

nomad edu report <group_name> [options]

Options:

Option	Description
`--db PATH`	Database path
`--days N`	Lookback period (default: 90)
`--json`	Output as JSON

Example output:

  NØMAÐ Group Report — cs101
  Last 90 days · 4 members · 4 ran jobs · 4 scored · 935 of 935 jobs measured · demo-cluster

  Median overall: 72/100 across 4 people (middle half 69–73)
  Over the period: 1 of 4 improved, 3 steady, 0 declined

  By dimension (median, people):
    CPU        51   (4)
    Memory     87   (4)
    Time       81   (4)
    I/O        69   (4)
    GPU        70   (4)

  Most common to work on:
    CPU: 4 of 4
    I/O: 2 of 4

  Member             Jobs  Measured  Overall  Change  Weakest
  diana               200       200       66      +1  I/O
  charlie             243       243       70      +9  CPU
  alice               260       260       73      +2  CPU
  bob                 232       232       73      -2  CPU

Every figure says how many people and jobs it rests on. A job counts when it finished in the period and is scored only when NØMAÐ measured it; members with no jobs are listed apart, never averaged in as zero. Figures over people are medians. "Change" compares a member's first and last week with measured jobs, and needs two such weeks. The Console's Group Reports page shows the same numbers.

`nomad edu me`¶

Your own jobs over the last 90 days (--days N), scored by the same rules as nomad edu report: every finished job is counted, only jobs NØMAÐ measured are scored, the overall score is the mean of your dimension averages, and the trend compares your first and last week with measured jobs. Your line in a group report and this page therefore show the same numbers, and so do the Console's Trajectory and My Activity pages. With --detailed:

  Your NØMAÐ Profile — alice
  ────────────────────────────────────────────────────────
  25 jobs in the last 90 days, 15 measured and scored
  Overall score: 52 / 100  (improving, +6 since your first week)
  By dimension (average over measured jobs): CPU 43, memory 62, time 50

  Top issues across your measured jobs:
  ────────────────────────────────────────────────────────

  [HIGH]     CPU Efficiency (spydur/basic) — ↑ improving
    5/7 jobs scored below threshold (avg score: 21.8)

    Your jobs:    request 8 (typical)
                  use 2.0 median (range 1.0–3.0 cores)
                  that's 25.0% utilization

    Try:          #SBATCH --ntasks=2

An issue is listed when most of your measured jobs on a cluster and partition fall below the threshold for that dimension. --detailed adds the line with your score in each dimension; --json gives everything as JSON; admins can pass --user. With jobs but none measured, it says so instead of showing a score.

Each suggested value says where it comes from: for memory and time, how many of the flagged jobs it fits ("fits what each of the 5 flagged jobs used, with a 2x safety buffer"; for one job, "fits what the flagged job used"); for cores and GPUs, how many flagged jobs needed it, from what they used.

Workstation sessions¶

Your workstation sessions of the last 7 days can add one piece of advice, listed first:

Memory: two or more sessions that used at least 80% of the workstation's memory for 12 hours or more, or ten or more that used that much at any length. Depending on the cluster's nodes it says to move to the cluster (a node with at least twice the peak), that the workstation is the right tool, or that the work is too big for both.
Long computation: sessions of 12 hours or more that kept at least half the workstation's cores (and at least 2) busy on average, adding up to a day or more. One long run counts. It says to run such work on the cluster, with a batch script sized from the session: cores it kept busy, twice its memory peak, and 1.5 times its length. A site without a cluster gets no such advice.

Setting Up Groups¶

NØMAÐ uses Linux groups for course/lab membership. To track a class:

Option 1: Use existing Linux groups¶

If your users are already in groups (e.g., bio301, cs101):

# Collect group membership
nomad collect -C groups --once

# Generate report
nomad edu report bio301

Option 2: Create dedicated groups¶

# Create group for course
sudo groupadd cs301

# Add students
sudo usermod -aG cs301 student01
sudo usermod -aG cs301 student02
# ...

# Collect and report
nomad collect -C groups --once
nomad edu report cs301

Option 3: Manual group file¶

Create a CSV file and import:

username,group_name,gid,cluster
alice,cs301,3001,spydur
bob,cs301,3001,spydur

nomad edu import-groups groups.csv

Dashboard Integration¶

The dashboard includes an Education tab showing:

Class-wide proficiency distributions
Individual student progress
Common problem areas
Improvement trends over time

Access via: nomad dashboard → Education tab

Best Practices¶

For Instructors¶

Baseline early: Collect data from the first week to establish starting points
Check weekly: Review group reports to identify struggling students early
Focus on trends: Individual job scores vary; trajectories matter more
Share reports: Let students see class-wide (anonymized) progress

For Mentors¶

Onboarding checkpoint: Review trajectory after first 10 jobs
Specific feedback: Use explain output to guide discussions
Celebrate improvement: Recognize when dimensions improve

For Users¶

Review failed jobs: Use explain to understand what went wrong
Track your trajectory: Check weekly to see improvement
Act on recommendations: The suggestions are data-driven

Technical Details¶

For detailed information on how proficiency is computed:

Proficiency Scoring — Formulas, dimensions, and scoring rubrics
Database Schema — How scores are stored

Troubleshooting¶

"Job not found in database"¶

Job 12345 not found in database.

Hint: Specify a database with --db or run 'nomad init' to configure.
  Example: nomad edu explain 12345 --db ~/nomad_demo.db

Solutions:

Specify the database: nomad edu explain 12345 --db /path/to/db
Run nomad init to configure the default database
Ensure data collection is running: nomad collect

"Not enough data for user"¶

The user needs at least 3 completed jobs for trajectory analysis.

"No data found for group"¶

Ensure group membership data has been collected:

nomad collect -C groups --once

Educational Analytics¶

Overview¶

Quick Start¶

Commands¶

nomad edu explain¶

nomad edu trajectory¶

nomad edu report¶

nomad edu me¶

Workstation sessions¶

Setting Up Groups¶

Option 1: Use existing Linux groups¶

Option 2: Create dedicated groups¶

Option 3: Manual group file¶

Dashboard Integration¶

Best Practices¶

For Instructors¶

For Mentors¶

For Users¶

Technical Details¶

Troubleshooting¶

"Job not found in database"¶

"Not enough data for user"¶

"No data found for group"¶

`nomad edu explain`¶

`nomad edu trajectory`¶

`nomad edu report`¶

`nomad edu me`¶