jitter
A random amount added to or taken from a retry delay so that many clients do not all retry at the same moment.
In more detail
Without jitter, clients that failed together retry together and hit the recovering service in synchronised waves. Full jitter picks a random delay between zero and the current backoff limit.
Related terms
- backoff — Waiting longer and longer between successive retries, often doubling the delay each time, so a struggling service has room to recover.
- retry — Sending a request again after it failed or timed out, in the hope that it succeeds the second time.
- thundering herd — A flood of requests that arrives at once when many clients react to the same event, such as all retrying after an outage.
- exponential backoff — Waiting progressively longer between retries, for example 1, 2, then 4 seconds, so a struggling service gets room to recover.
- retry storm — A flood of retries from many clients at once that hits an already struggling service harder than the original traffic.
- mass expiry — Many cache keys expiring at the same moment because they were written together with the same time to live.
Taught in
- When The Call Fails — AI Engineering: LLM Apps, RAG & Agents
- Error Handling And Safe Retries — Modern API Architecture: REST, GraphQL & gRPC
- Rate Limiting Strategies — Modern API Architecture: REST, GraphQL & gRPC
- Built-in Rate Limiting Middleware — Modern .NET Core, C# & Enterprise Microservices
- Retry and Circuit Breaker Policies with Polly — Modern .NET Core, C# & Enterprise Microservices
- Bidirectional WebSockets: Reconnection & Heartbeats — Production Full-Stack Engineering
- Cache Invalidation And The Three Hard Cases — Scalable System Design & Distributed Architecture
- Retries, Backoff And Jitter — Scalable System Design & Distributed Architecture
References
- Amazon Builders’ Library: Timeouts, retries and backoff with jitter
- Polly: retry strategy (jitter)
- Transient fault handling
- AWS Architecture Blog: Exponential backoff and jitter