Tail latency reduction by racing redundant requests.
Starts the given function in a Task, waits for :delay milliseconds,
and if no result has arrived, fires up to :max_hedged additional
Tasks running the same function. The first successful result wins.
This is useful when a backend has variable latency and you would rather pay the cost of a redundant request than wait for a slow one.
Options
:name-- optional atom for telemetry metadata. Default:hedge.:delay-- milliseconds to wait before firing hedge requests. Required.:max_hedged-- max additional requests to fire. Default1.
Examples
iex> ExResilience.Hedge.call(fn -> {:ok, 42} end, delay: 100)
{:ok, 42}Summary
Functions
Executes fun with hedge logic for tail latency reduction.
Types
@type option() :: {:name, atom()} | {:delay, non_neg_integer()} | {:max_hedged, pos_integer()}
Functions
Executes fun with hedge logic for tail latency reduction.
Starts fun as the primary request. If it does not complete within
:delay milliseconds, fires up to :max_hedged additional copies.
Returns the first successful result, or the first error if all fail.
Examples
iex> ExResilience.Hedge.call(fn -> {:ok, :fast} end, delay: 50)
{:ok, :fast}