Backoff

Pure calculation helper for exponential backoff delays. Instead of hardcoding delay arrays or hand-rolling Math.Pow in every retry loop, create a BackoffPolicy once and call Delay(attempt).

Usage

using Servus.Resilience;

private static readonly BackoffPolicy _backoff =
    Backoff.Create(TimeSpan.FromMilliseconds(500), maxDelay: TimeSpan.FromSeconds(30));

// Wherever you need the delay:
var delay = _backoff.Delay(attempt);
// 500ms → 1s → 2s → 4s → …

With jitter

Jitter prevents multiple callers from retrying in lock-step (thundering herd):

var delay = _backoff.DelayWithJitter(attempt);

With Akka.NET timers

private static readonly BackoffPolicy _backoff =
    Backoff.Create(TimeSpan.FromMilliseconds(500), maxDelay: TimeSpan.FromSeconds(30));

private void HandleQueueFull(int attempt)
{
    if (attempt < MaxRetries)
    {
        Timers.StartSingleTimer("retry", new RetryQuery(attempt + 1), _backoff.Delay(attempt));
    }
}

Semantics

• attempt is zero-based — attempt 0 returns initialDelay.
• The delay for attempt n is initialDelay × multiplier^n.
• When the calculated delay exceeds maxDelay, maxDelay is returned.
• Overflow from very high attempt counts is handled safely (caps at maxDelay).
• DelayWithJitter multiplies the base delay by a random factor in [0.75, 1.25).

Default progression (500 ms initial, 2× multiplier)

Attempt	Delay
0	500 ms
1	1 s
2	2 s
3	4 s
4	8 s
5	16 s

API

public static class Backoff
{
    public static BackoffPolicy Create(
        TimeSpan initialDelay,
        double multiplier = 2.0,
        TimeSpan? maxDelay = null);
}

public sealed class BackoffPolicy
{
    public TimeSpan Delay(int attempt);
    public TimeSpan DelayWithJitter(int attempt);
}

When to use it

• Actor retry timers — compute the next delay for Timers.StartSingleTimer.
• Manual retry loops — anywhere you'd otherwise hardcode a TimeSpan[] of backoff values.
• Queue-full / rate-limit handling — back off before retrying a request.

When NOT to use it

• You need a full retry policy with circuit breakers, bulkheads, etc. — use Polly.
• You want an all-in-one "call this function with retries" helper — wrap the policy in your own loop or use Polly's WaitAndRetryAsync.