Fill 0001 — Backoff and Jitter

A compact, testable model of retry delays, with a CLI to simulate schedules.

Code: lcrc-experiments/fills/0001-backoff-jitter

Accelerator

Retries are a reliability tool, and also a reliability hazard. This fill models exponential backoff with jitter, so retries remain controlled under load.

MAPS

• Input: retry attempt number (0..N), policy parameters (base, factor, cap, jitter).
• Output: a delay value (seconds).
• Deliverable: a small Python package + CLI + unit tests.

BOB

References worth reading:

• AWS: Exponential Backoff and Jitter
• Google SRE Book — Handling Overload
• Google Cloud — Retry strategy / exponential backoff

CMS

CLI example:

lcrc-backoff --max-attempts 6 --base 0.5 --cap 10 --seed 42 --json

LCRCb

• Determinism: fixed seed must reproduce schedules.
• Bounds: jittered delays must remain within expected ranges.
• Cap: exponential growth must stop at the cap.