Question

To avoid the thundering herd problem during system recovery, what specific technique should be applied to retry logic in addition to exponential backoff?

Accepted Answer

To avoid the thundering herd problem when using exponential backoff, you must apply jitter. The thundering herd problem occurs when many clients that have failed to connect all retry their requests at the exact same moment after a delay, causing a sudden spike in traffic that crashes the system again. Exponential backoff is a strategy where a client increases the wait time between each consecutive failed attempt, but if multiple clients start this process at the same time, they will remain synchronized and continue to retry in unison. Jitter is the addition of random variation to the retry delay. By introducing randomness, each client calculates a slightly different wait time, which causes the retry attempts to spread out over a wider window of time. Instead of hitting the server in a single, massive wave, the requests are distributed evenly, allowing the system to recover gracefully without being overwhelmed by a synchronized surge of traffic.

To avoid the thundering herd problem during system recovery, what specific technique should be applied to retry logic in addition to exponential backoff?

To avoid the thundering herd problem when using exponential backoff, you must apply jitter. The thundering herd problem occurs when many clients that have failed to connect all retry their requests at the exact same moment after a delay, causing a sudden spike in traffic that crashes th....

Community Answers

Home → All Courses → Engineering and Technology Courses → Distributed Systems for Internet-Scale Services → Flashcard

To avoid the thundering herd problem during system recovery, what specific technique should be applied to retry logic in addition to exponential backoff?

To avoid the thundering herd problem when using exponential backoff, you must apply jitter. The thundering herd problem occurs when many clients that have failed to connect all retry their requests at the exact same moment after a delay, causing a sudden spike in traffic that crashes th....

Community Answers