Why include jitter with exponential backoff?

Why include jitter with exponential backoff?

• A. To ensure uniform wait times across all clients.
• B. To avoid synchronized retries that create retry storms.
• C. To guarantee immediate success on retry.
• D. To bypass the need for exponential waiting.

Answer: (B)

Jitter with exponential backoff desynchronizes retries across many clients that experienced a failure, so they don’t all come back at the same moment. Exponential backoff already reduces load by increasing wait times after each failure, but without randomness, a lot of clients can still retry in lockstep when the service recovers. Adding jitter randomizes each client’s delay within a range, spreading the retry attempts over time and preventing a sudden surge that could overwhelm the system again. This keeps the service more stable and improves overall availability during recovery. Uniform wait times would still trigger synchronized retries and retry storms. Expecting immediate success on retry isn’t guaranteed and defeats the purpose of backoff. Skipping exponential waiting defeats the very reason backoff exists, which is to throttle retries and reduce overload.

Jitter with exponential backoff desynchronizes retries across many clients that experienced a failure, so they don’t all come back at the same moment. Exponential backoff already reduces load by increasing wait times after each failure, but without randomness, a lot of clients can still retry in lockstep when the service recovers. Adding jitter randomizes each client’s delay within a range, spreading the retry attempts over time and preventing a sudden surge that could overwhelm the system again. This keeps the service more stable and improves overall availability during recovery.

Uniform wait times would still trigger synchronized retries and retry storms. Expecting immediate success on retry isn’t guaranteed and defeats the purpose of backoff. Skipping exponential waiting defeats the very reason backoff exists, which is to throttle retries and reduce overload.