ReferenceGlossary
Hedged request
Sending a second copy of a request to another replica when the first is slower than expected, for example past the 95th-percentile latency, and using whichever answer arrives first. It cuts tail latency for a small amount of extra load.
Performance
In Quorum#
Not modeled yet: no Link or research sends a hedged copy. The closest is the Wide-column Store, where a read waits only for the fastest R of its N replicas, so a lower R gives a faster read. Percentiles shows the p99 that hedging would cut.
Read next#
Tail at scale, Percentile (p50, p95, p99), Quorum, Retry with backoff and jitter