Pulse Documentation

name: op-synth-exponential
description: Exponential samples with rate parameter lambda; mean is 1/lambda.
kind: operator
category: SYNTH
operator: exponential
type: reference
applies_to: inspect, predict, manifest
examples_tags: [synth, time-series, distribution-shape]

Synth distributions emit per-row values; no Response.Components.

Use when

Draws positive waiting times where short waits are common and long ones rare, set by how often events happen.

Questions it answers:

• How do I simulate the time between customer arrivals?
• How do I generate realistic call durations for a load test?

Use something else:

• poisson when you want whole-number counts of events per period.
• pareto when a few values should be extremely large.

Params

Inputs

Output

Per-row float64 sample from Exp(lambda), computed as rng.ExpFloat64() / lambda. Tail is unbounded — heavy enough to inflate u4/u8 overflow risk under low lambda.

Gotchas

• lambda <= 0 → SERVICE_VALIDATION at spec parse.
• Memoryless property: hazard rate is constant. Pair with monotonic_from for inter-arrival timestamps (cumulative sum captured downstream via a window operator).
• No clamp param — bound via constraints (e.g. delay_s < 86400) if a runaway tail breaks the host field’s cast.
• Common pattern: exponential lifetimes / wait-times, log-transformed for distribution-shape work.

See

• pulse_examples_search tags=[synth]
• Skills: synthetic-data, op-synth-poisson, op-synth-pareto