Fibonaccibackoff

Fibonaccibackoff is a retry backoff strategy used in distributed systems. It uses the Fibonacci sequence to determine the delay between retry attempts, typically by mapping the n-th retry to a delay proportional to Fibonacci(n) units of time. The approach blends the idea of a Fibonacci backoff with practical constraints such as maximum delay and jitter to avoid thundering herd problems.

Origins: The term appears in academic and practitioner writings as a hypothetical or coined strategy blending Fibonacci growth with backoff principles. It is not as widely standardized as exponential backoff; implementations vary.

Mechanism: You maintain a retry counter n. The base delay is some time unit (e.g., milliseconds). The delay before the next retry is min(Fib(n) * base, maxDelay). Jitter is added by randomizing within a range to reduce contention. Optional warm-up and cap mechanisms exist to prevent excessive delays.

Variants: Some variants apply Fibonacci to only certain ranges of retries, or use Fibonacci plus a linear or exponential cap; some implementations use scaled Fibonacci numbers with a factor to reduce or amplify delays.

Use cases: When simple incremental delays are desired without the rapid growth of exponential backoff, e.g., systems with moderate backpressure or cost to retries; IoT devices; API clients; message queuing.

Comparison: Compared with exponential backoff, Fibonaccibackoff grows more slowly early on and more rapidly later, but typically slower to reach long delays than standard exponential. It can offer a middle ground between immediate retry and long wait times.

See also: Backoff algorithm, Fibonacci sequence, jitter, retry policy.