Immediate retries vs exponential backoff
Immediate retries send a failed request again at once, so 1,000 callers/s add a second wave to a service that tries to recover. Exponential backoff waits longer after each failure. After a 2 s link cut, immediate retries leave 15 percent of requests failed. Backoff leaves none.
3 retries per call · no jitter · peak 6.0× normal load
Case 1
1,000 requests/s reach a one-core service. The link fails from 0:02 to 0:04. At least 99.9 percent of requests must succeed, with p99 below 5 s. Immediate retries give p99 7,936 ms and backoff 4,032 ms.
| Measure | Goal | 5 retries with no wait | Backoff from 750 ms, with jitter |
|---|---|---|---|
| Error rate | under 0.1% | 15% Fail | 0% Pass |
| p99 latency | under 5 s | 7.9 s Fail | 4 s Pass |
| Monthly cost | 4 USD/mo | 4 USD/mo |