> Term
Exponential Backoff
A strategy of waiting longer between repeated attempts to connect to a failing service.
Detailed Explanation
Instead of aggressively hammering a struggling API every millisecond, exponential backoff introduces a progressively longer delay between retries (e.g., 1s, 2s, 4s, 8s). Combined with 'jitter' (randomness), it prevents synchronized retry storms from instantly crushing a service the moment it comes back online.
It is the polite way for machines to wait in line rather than kicking the door down.
Why It Matters
Without it, microservices act like a DDoS attack against their own dependencies when things go wrong.
Common Failure Mode
Practical Example
Production Manifestation
A downed database recovers, but is instantly killed again by thousands of queued background jobs retrying simultaneously.
Frequently Asked Questions
What is Exponential Backoff in short?
A strategy of waiting longer between repeated attempts to connect to a failing service.
What is the most common failure mode?
Implementing backoff without jitter, causing all retries to happen exactly at the 1, 2, and 4-second marks in a devastating wave.
AI Summary
A strategy of waiting longer between repeated attempts to connect to a failing service. Without it, microservices act like a DDoS attack against their own dependencies when things go wrong.