Dedicated read/write stress (closed-loop workers) go run ./bench -stress
Ceiling / saturation studies (open-loop dispatch + per-class queues) go run ./bench -ceiling
Flags
-matrix — run matrix mode (profiles x worker grid)
-sweep — run read-only sweep mode
-stress — run dedicated read/write stress mode
-ceiling — run ceiling mode with class dispatchers/queues
-case-duration — case duration for matrix and stress
-sweep-duration — case duration for sweep
-ceiling-stage-duration — stage duration for ceiling
-report-every — progress log interval
-out — output JSON report path
-email-sample — number of sampled emails for single-record read scenario
-stress-no-read — disable read workers in stress
-stress-no-write — disable write workers in stress
-ceiling-dispatch-quantum — dispatcher quantum for offered load generation
-ceiling-queue-cap — per-class queue capacity
-ceiling-suites — all or list: read_fast_ceiling,write_update_ceiling,write_insert_ceiling,read_fast_under_write,slow_vs_fast,mixed_fast_slow_write,low_write_regression
-ceiling-read-fast-grid — target ops/s grid for fast-read ceiling
-ceiling-write-grid — target ops/s grid for write ceiling/interference suites
-ceiling-read-slow-grid — target ops/s grid for slow-read interference
-ceiling-low-write-grid — low-write grid for early-regression sweep
-ceiling-read-fast-workers / -ceiling-read-slow-workers — executors for read classes
-ceiling-write-update-workers / -ceiling-write-insert-workers — executors for write classes
-ceiling-mixed-slow-ops — fixed slow-read offered load in mixed suite
-ceiling-baseline-fraction — fraction of discovered fast-read ceiling used as baseline