Performance

Time Wheel Benchmarks

Apple M4 / arm64. Pure scheduling engine, no Redis.

Test	Result
1M Add	225ms (4.4M tasks/sec, 225 ns/op)
1M Add + Fire	1.2s (all fired, zero loss)
4 writers x 250K	187ms (5.3M tasks/sec)
1M Cancel	268ms (290 ns/op)

vs Redis Sorted Set

Operation	seqdelay (time wheel)	Redis ZRANGEBYSCORE
Add	225 ns (local memory)	~0.5 ms (network RTT)
Fire check	O(1) per tick	O(log N) per poll
Precision	1ms configurable	~1s (poll interval)
Redis load	On state change only	Every poll cycle

Running Benchmarks

# Time wheel only (no Redis needed)
go test -bench=BenchmarkTimeWheel -benchmem -count=1

# Stress test: 1M tasks
go test -run TestStress -v -timeout 120s

# Full lifecycle (needs Redis)
go test -bench=BenchmarkQueue -benchmem -timeout 60s

# All benchmarks
go test -bench=. -benchmem -count=1 -timeout 120s