Retry with "hedge" to reduce tail latency at scale #1050
Description
Hello,
Grafana dskit uses "hedging" with "cristalhq/hedgedhttp" in order to reduce latency when quering endpoints.
- https://github.com/cristalhq/hedgedhttp
- https://github.com/grafana/dskit/tree/main/hedging
- https://www.barroso.org/publications/TheTailAtScale.pdf
hedgedhttp said
See paper Tail at Scale by Jeffrey Dean, Luiz André Barroso. In short: the client first sends one request, but then sends an additional request after a timeout if the previous hasn't returned an answer in the expected time. The client cancels remaining requests once the first result is received.
Do you think that it's could be an accepted feature ?