Struct GroupCommitBuffer 

pub struct GroupCommitBuffer { /* private fields */ }

Group commit buffer for batching WAL writes

Reduces fsync overhead by batching multiple transactions. Uses Little’s Law for adaptive batch sizing: N* = sqrt(2 × L_fsync × λ / C_wait)

Implementations

impl GroupCommitBuffer

pub fn new(max_pending: usize, max_wait_us: u64) -> Self

Create new group commit buffer

pub fn with_fsync_latency( max_pending: usize, max_wait_us: u64, fsync_latency_us: u64, ) -> Self

Create with custom fsync latency estimate

pub fn add(&self, txn_id: u64) -> bool

Add a transaction to pending commits

Returns true if buffer should be flushed.

pub fn take_pending(&self) -> Vec<PendingCommit>

Take pending commits for flush

pub fn record_fsync_latency(&self, latency_us: u64)

Record actual fsync latency for calibration

pub fn current_batch_size(&self) -> usize

Get current adaptive batch size

pub fn current_arrival_rate(&self) -> f64

Get current arrival rate estimate (req/s)

pub fn stats(&self) -> GroupCommitStats

Get statistics for monitoring