Struct LeakBucket

pub struct LeakBucket {
    pub(crate) scale: AtomicU64,
    pub(crate) ltat: AtomicU64,
    pub(crate) origin: Instant,
    pub(crate) total_allowed_quotas: AtomicU64,
    pub(crate) total_waited_nanos: AtomicI64,
}

A GCRA-like leak bucket visual scheduler that never deny request even whose weight is larger than tau and only count TAT.

Fields

scale: AtomicU64

Weight scale per 1.0 unit quota in nanosecond.

scale is always non-zero.

scale = rate / 1 (in second)

ltat: AtomicU64

Last request’s TAT (Theoretical Arrival Time) in nanosecond.

origin: Instant

Zero time instant.

total_allowed_quotas: AtomicU64

Total allowed quotas.

total_waited_nanos: AtomicI64

Total waited nanos.

Implementations

impl LeakBucket

pub(crate) const NANO: u64 = 1_000_000_000u64

pub(crate) fn scale(rate: NonZeroU64) -> u64

calculate the weight scale per 1.0 unit quota in nanosecond.

pub(crate) fn new(rate: NonZeroU64) -> Self

Create a new GCRA-like leak bucket visual scheduler with given rate.

pub(crate) fn check(&self, quota: u64) -> Result<(), Duration>

Check if the request with the given quota is supposed to be allowed at the moment.

On success, the quota will be consumed. The caller is supposed to proceed the quota.

On failure, the minimal duration to retry check() is returned.

pub(crate) fn _avg_wait_nanos_per_quota(&self) -> i64

Average wait time per quota.

Positive value indicates waits, negative value indicates there is spare rate limit.

pub(crate) fn _reset_stats(&self)

Reset statistics.

pub(crate) fn _update(&self, rate: NonZeroU64)

Update rate limit with the given rate.

Trait Implementations

impl Debug for LeakBucket

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

Formats the value using the given formatter.