Per-stage latency breakdown (bench-trace)
Mean µs per stage, averaged over all runs. Captured with
--features bench-trace. queue = enqueue→worker-flush; bridge_wait = worker-flush→flusher-dequeue; write = flusher-dequeue→write_record (pre-fsync); total = enqueue→ack-fired.
deadline = d1ms
| Stage | stage_sync_d1ms | stage_batched_d1ms | stage_buffered_d1ms |
|---|---|---|---|
| Queue (µs) | 2 | 2 | 3 |
| Bridge wait (µs) | 4 | 4 | 7 |
| Write (µs) | 12 | 11 | 8 |
| Total (µs) | 166 | 152 | 18 |
deadline = d2ms
| Stage | stage_sync_d2ms | stage_batched_d2ms | stage_buffered_d2ms |
|---|---|---|---|
| Queue (µs) | 2 | 2 | 3 |
| Bridge wait (µs) | 4 | 4 | 8 |
| Write (µs) | 9 | 9 | 6 |
| Total (µs) | 158 | 145 | 18 |