Parameter Sweep

The Sweep tab in the Overview panel runs a Monte Carlo simulation across a grid of parameter values and renders the results as a color-coded heatmap. It’s the fastest way to find the right capacity, service rate, or split weight for your design without manually tweaking and re-running.

When to use it

You’ve simulated your design and you know there’s a bottleneck. You want to know:

• “What capacity does this queue need to stop dropping?”
• “How fast does this worker have to be before the upstream backs up?”
• “What split weight should this router have so neither path saturates?”
• “If both capacity AND service rate move together, what’s the cheapest combo that keeps health above 90?”

A parameter sweep answers all of these in one run.

How to set it up

1. Open the Overview panel (O).
2. Click the Sweep tab.
3. Configure the X axis:
   • Component — pick which block to vary.
   • Knob — capacity, service rate, or split weight (for the first weighted outbound).
   • Min / Max / Steps — the value range and how many cells along the axis (2-10).
4. Optionally enable the Y axis to make it a 2D sweep (same fields).
5. Set Runs/cell — how many Monte Carlo replications per grid cell. More = tighter confidence intervals, slower run. Default 5.
6. Hit Run sweep.

Total simulation count = X steps × Y steps × Runs/cell. A 5×5 grid at 5 runs/cell = 125 sims. The float-ms WASM engine usually finishes a sweep that size in seconds.

Reading the heatmap

Each cell shows the value of one of four metrics:

• Throughput (delivered) — mean number of items that reached terminal sinks. Higher is better.
• Dropped — mean number of items dropped (queue overflow, filter rejects, retries exhausted). Lower is better.
• Peak utilization — max ρ observed in any component during the run. Lower is better (high peak = local hotspot).
• Mean health — mean health score across runs (0-100). Higher is better.

Toggle the metric with the Show: dropdown. Color tiers match the canvas sparklines:

• Green — value is in the best 25% of the range.
• Lime — best 50%.
• Amber — worst 50%.
• Red — worst 25%.

Color polarity flips automatically for lower-is-better metrics (dropped, peak utilization). Hover any cell for the full tooltip with all four metrics.

Knobs explained

Capacity

The queue depth on the target component. Sweeping capacity tells you the smallest cap that doesn’t drop under your stress scenario.

Service rate

Parallel processing slots on the target component (e.g. 1 worker, 2 baristas, 5 servers). Sweeping service rate tells you the staffing level needed to keep the queue from backing up.

Split weight (1st outbound)

The weight on the first weighted outbound of a split-mode component. The remaining outbounds are renormalized so the total stays at 100. Sweeping weight tells you how routing ratio affects load distribution downstream.

Examples

1D sweep, finding minimum capacity:

• X axis: Pad Thai Path, capacity, min 5, max 50, steps 6
• Y axis: off
• Metric: dropped

You’ll see at which capacity the dropped count hits 0. Lowest capacity that’s all-green is the minimum to provision.

2D sweep, capacity vs service rate:

• X axis: Kitchen Cooking, capacity, 5–50, 6 steps
• Y axis: Kitchen Cooking, service rate, 1–10, 5 steps
• Metric: mean health

The heatmap shows the contour where health crosses 90. The cheapest point on that contour is the recommended config. Save the baseline before changing the design so you can compare.

1D sweep on routing weight:

• X axis: Order Method, split weight, 0–100, 11 steps
• Y axis: off
• Metric: peak utilization

You’ll see how the bottleneck shifts as you reroute traffic. The sweet spot is the weight where peak utilization is minimized (load is evenly distributed).

Limits

• 2-10 steps per axis (sweeping more than 10 cells per axis usually means you should narrow the range first).
• 1-20 runs per cell.
• Sweep is a Pro feature.

How it relates to Monte Carlo

Monte Carlo runs N replications of the same canvas with varied seeds and tells you the variance in your single design. A parameter sweep runs Monte Carlo at every cell in a grid and tells you how the metrics change as a knob moves. Use Monte Carlo when you want to know “how stable is THIS design”; use sweep when you want to know “what setting should I pick”.