Retry and backoff simulator
Read a Retry-After header, then see what happens when many clients hit a rate limit at the same moment and retry with different strategies. Runs in your browser; we only count anonymous visits.
1. Retry-After header
2. Many clients, one limit
How to use this tool
- Paste the
Retry-Aftervalue from a 429 or 503 response: a number of seconds or an HTTP-date in GMT. - Describe your load: how many clients fail together and how many requests per second the server can take.
- Compare the strategies in the table: requests sent in total (wasted work), the busiest second of retries and the time until every client succeeded.
Example
With 200 clients and 20 accepted per second, fixed delays keep every client in step, so each wave hits the limit again. Plain exponential backoff spreads waves apart but leaves clients waiting far longer than needed. Adding jitter lowers both the wasted requests and the total time in the default scenario.
Limits
This is a simplified model: all clients start together, the server counts whole seconds, and requests need no processing time. It shows the shape of the effect, not your real numbers. If the server sends Retry-After, clients should wait at least that long; this page does not apply it to the simulation unless you add it as the base delay.
Privacy and security of what you enter
Everything runs in your browser; nothing is uploaded. See the privacy notice.
We review rate-limit policies, timeouts and client retry behaviour. See the services, the HTTP status codes guide or the capacity calculator.
Message us on LinkedIn