Skip to content

Analytics Overview

Analytics is where you stop describing your system and start interrogating it. The simulator already runs your design step by step (see Runtime → Engine Methodology for how that works). Analytics is the layer that turns runtime data into answers.

Question you’re askingOpen this
”Does my design even work under expected load?”Stress Testing
”Where is my latency budget going?”Timeline View (Pro)
“How does this design hold up across many random seeds?”Monte Carlo (Pro)
“Is health stable over time, or drifting?”Stability Analysis (Pro)
“What’s the right value for this capacity / service rate?”Parameter Sweep (Pro)

Stress Testing is the foundation. Every other analytics view runs on top of the same deterministic engine — they just slice the resulting data differently:

  • Stress Testing — single run, single seed, you manually adjust load / faults / latency. Read the bottleneck, fix, re-run. Free for everyone.
  • Timeline View — same run, re-rendered along a time axis. The longest bar IS the bottleneck. Spine sub-mode for storytelling, Gantt sub-mode for diagnosing. Pro.
  • Monte Carlo — same design, run N times with varied seeds, aggregate the distribution. Answers “is my P95 healthy, not just my median?” Pro.
  • Stability Analysis — same single run, scored frame-by-frame for consistency. Catches drift / oscillation that average metrics miss. Pro.
  • Parameter Sweep — Monte Carlo across a 1D or 2D grid of behavior knobs. Heatmap output. Answers “what’s the optimal value?” instead of “is THIS value good?” Pro.

The typical analytics loop:

  1. Stress test first. Hit your design with realistic load. Read the bottleneck.
  2. Open Timeline View to see where the time goes. Critical path highlighted in red.
  3. Fix the worst offender in Topology view. Bump capacity, add a queue, swap behavior.
  4. Re-stress. Verify the fix actually moved the bottleneck.
  5. Once it looks good on one seed: open Monte Carlo to confirm it holds across variation.
  6. If a metric is sensitive to a single knob: Parameter Sweep finds the right value without guesswork.
  7. For long-running systems: check Stability Analysis to make sure health is consistent over time, not just averaging out to “okay.”

Each tool answers a different question. Don’t reach for Monte Carlo when you haven’t run Stress Testing once. Don’t run Parameter Sweep blind — let Stress Testing tell you which knob matters first.