Retry Strategies: Exponential Backoff & Jitter

In distributed systems, transient network glitches, TCP resets, and brief CPU spikes are inevitable. A naive implementation attempts to handle network errors by immediately retrying the failed request:

typescript
Copy
// ❌ NAIVE IMMEDIATE RETRY ANTI-PATTERN
async function naiveFetch(url: string, retries = 3) {
  for (let i = 0; i < retries; i++) {
    try {
      return await http.get(url);
    } catch (err) {
      if (i === retries - 1) throw err;
      // Immediate retry with 0ms delay!
    }
  }
}

Why This Destroys Backend Infrastructure:

1. A backend database experiences a momentary 200ms lock stall.
2. 5,000 in-flight client requests time out.
3. All 5,000 clients immediately retry at the exact same millisecond.
4. The incoming traffic instantly doubles to 10,000 QPS on a server that was already struggling.
5. The Retry Storm: A minor transient hiccup escalates into a complete, permanent system outage as cascading retries overwhelm the server.