Struct Aimd 

pub struct Aimd { /* private fields */ }

AIMD (Additive Increase / Multiplicative Decrease) concurrency limit strategy.

A loss-based algorithm that increases the limit by a fixed amount on success and multiplies it by a backoff ratio on error or timeout. This is the same approach used in TCP Reno congestion control.

Unlike Vegas, AIMD does not track baseline RTT or estimate queue depth — it reacts purely to errors and timeouts, making it simpler but less proactive.

Differences from Netflix’s Java implementation

The Java reference truncates the limit to an integer after every update, while this implementation keeps it as an f64 internally. This means repeated backoffs decay more smoothly (e.g. 10.0 → 9.0 → 8.1 → 7.29 instead of 10 → 9 → 8 → 7). The observable limit (via max_concurrency) is the same in most cases, but after recovery the internal state may be slightly higher than the Java equivalent, leading to marginally faster ramp-up.

Implementations

impl Aimd

pub fn builder() -> AimdBuilder

Returns an AimdBuilder for configuring a new Aimd instance.

Trait Implementations

impl Algorithm for Aimd

fn max_concurrency(&self) -> usize

Returns the maximum number of concurrent requests the algorithm currently allows.

fn update( &mut self, rtt: Duration, num_inflight: usize, is_error: bool, is_canceled: bool, )

Observes the outcome of a request and updates the algorithm’s state accordingly.

impl Clone for Aimd

fn clone(&self) -> Aimd

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for Aimd

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for Aimd

fn default() -> Self

Returns the “default value” for a type.