go-lock
go-lock is a Golang library implementing an effcient read-write lock with the following built-in mechanism:
- Mutex with timeout mechanism
- Trylock
- No-starve read-write solution
Native sync/Mutex and sync/RWMutex are very powerful and reliable. However, it became a disaster if the lock was not released as expected. Or, someone was holding the lock too long at the peak time leading whole system blocked. Dealing with those cases, go-lock implements TryLock, TryLockWithTimeout and TryLockWithContext function in addition to Lock and Unlock. It provides flexibility to control the resources.
Installation
go get github.com/viney-shih/go-lockExample
package main
import (
"fmt"
"sync/atomic"
"time"
lock "github.com/viney-shih/go-lock"
)
func main() {
// set RWMutex with CAS mechanism (CASMutex).
var rwMut lock.RWMutex = lock.NewCASMutex()
// set default value
count := int32(0)
// block here
rwMut.Lock()
go func() {
time.Sleep(50 * time.Millisecond)
fmt.Println("Now is", atomic.AddInt32(&count, 1)) // Now is 1
rwMut.Unlock()
}()
// waiting for previous goroutine releasing the lock, and locking it again
rwMut.Lock()
fmt.Println("Now is", atomic.AddInt32(&count, 2)) // Now is 3
// TryLock without blocking
// Return false, because the lock is not released.
fmt.Println("Return", rwMut.TryLock())
// RTryLockWithTimeout without blocking
// Return false, because the lock is not released.
fmt.Println("Return", rwMut.RTryLockWithTimeout(50*time.Millisecond))
// TryLockWithContext without blocking
ctx, cancel := context.WithTimeout(context.TODO(), 50*time.Millisecond)
defer cancel()
// Return false, because the lock is not released.
fmt.Println("Return", rwMut.TryLockWithContext(ctx))
// release the lock in the end.
rwMut.Unlock()
// Output:
// Now is 1
// Now is 3
// Return false
// Return false
// Return false
}Benchmarks
- Run on MacBook Pro (Retina, 15-inch, Mid 2015) 2.5 GHz Quad-Core Intel Core i7 16 GB 1600 MHz DDR3 using Go 1.15.2
- Run with 1, 2, 4, 8 and 16 cpu to show it scales well...16 is double the # of logical cores on this machine.
go test -cpu=1,2,4,8,16 -bench=. -benchmem=true
goos: darwin
goarch: amd64
pkg: github.com/viney-shih/go-lock
(sync.RWMutex)
BenchmarkRWMutexLock 42591765 27.4 ns/op 0 B/op 0 allocs/op
BenchmarkRWMutexLock-2 42713179 27.5 ns/op 0 B/op 0 allocs/op
BenchmarkRWMutexLock-4 44348323 27.5 ns/op 0 B/op 0 allocs/op
BenchmarkRWMutexLock-8 44135148 27.9 ns/op 0 B/op 0 allocs/op
BenchmarkRWMutexLock-16 43566165 27.5 ns/op 0 B/op 0 allocs/op
BenchmarkConcurrentRWMutexLock 42971319 27.5 ns/op 0 B/op 0 allocs/op
BenchmarkConcurrentRWMutexLock-2 20131005 57.5 ns/op 0 B/op 0 allocs/op
BenchmarkConcurrentRWMutexLock-4 10752337 115 ns/op 0 B/op 0 allocs/op
BenchmarkConcurrentRWMutexLock-8 11434335 105 ns/op 0 B/op 0 allocs/op
BenchmarkConcurrentRWMutexLock-16 10495626 109 ns/op 0 B/op 0 allocs/op
BenchmarkConcurrent50RWMutexLock 27979630 42.8 ns/op 0 B/op 0 allocs/op
BenchmarkConcurrent50RWMutexLock-2 13037742 86.6 ns/op 0 B/op 0 allocs/op
BenchmarkConcurrent50RWMutexLock-4 9143397 134 ns/op 0 B/op 0 allocs/op
BenchmarkConcurrent50RWMutexLock-8 8335652 139 ns/op 0 B/op 0 allocs/op
BenchmarkConcurrent50RWMutexLock-16 7876855 150 ns/op 0 B/op 0 allocs/op
(ChanMutex)
BenchmarkChanMutexLock 22619928 51.8 ns/op 0 B/op 0 allocs/op
BenchmarkChanMutexLock-2 22769630 51.6 ns/op 0 B/op 0 allocs/op
BenchmarkChanMutexLock-4 23096103 51.7 ns/op 0 B/op 0 allocs/op
BenchmarkChanMutexLock-8 22627267 51.1 ns/op 0 B/op 0 allocs/op
BenchmarkChanMutexLock-16 23092266 51.7 ns/op 0 B/op 0 allocs/op
BenchmarkConcurrentChanMutexLock 23422556 51.8 ns/op 0 B/op 0 allocs/op
BenchmarkConcurrentChanMutexLock-2 6101949 201 ns/op 0 B/op 0 allocs/op
BenchmarkConcurrentChanMutexLock-4 5882083 200 ns/op 0 B/op 0 allocs/op
BenchmarkConcurrentChanMutexLock-8 5827183 211 ns/op 0 B/op 0 allocs/op
BenchmarkConcurrentChanMutexLock-16 5577098 215 ns/op 0 B/op 0 allocs/op
BenchmarkChanMutexTryLock 22272500 51.8 ns/op 0 B/op 0 allocs/op
BenchmarkChanMutexTryLock-2 23004806 52.2 ns/op 0 B/op 0 allocs/op
BenchmarkChanMutexTryLock-4 22461870 52.3 ns/op 0 B/op 0 allocs/op
BenchmarkChanMutexTryLock-8 22901328 53.8 ns/op 0 B/op 0 allocs/op
BenchmarkChanMutexTryLock-16 22334739 55.4 ns/op 0 B/op 0 allocs/op
BenchmarkConcurrentChanMutexTryLock 22112250 53.9 ns/op 0 B/op 0 allocs/op
BenchmarkConcurrentChanMutexTryLock-2 13806072 85.0 ns/op 0 B/op 0 allocs/op
BenchmarkConcurrentChanMutexTryLock-4 45892635 32.6 ns/op 0 B/op 0 allocs/op
BenchmarkConcurrentChanMutexTryLock-8 396310569 2.86 ns/op 0 B/op 0 allocs/op
BenchmarkConcurrentChanMutexTryLock-16 512990590 2.35 ns/op 0 B/op 0 allocs/op
(CASMutex)
BenchmarkCASMutexLock 6160122 186 ns/op 96 B/op 1 allocs/op
BenchmarkCASMutexLock-2 7507022 156 ns/op 96 B/op 1 allocs/op
BenchmarkCASMutexLock-4 7646648 156 ns/op 96 B/op 1 allocs/op
BenchmarkCASMutexLock-8 7559616 156 ns/op 96 B/op 1 allocs/op
BenchmarkCASMutexLock-16 7576237 158 ns/op 96 B/op 1 allocs/op
BenchmarkConcurrentCASMutexLock 6375879 185 ns/op 96 B/op 1 allocs/op
BenchmarkConcurrentCASMutexLock-2 1690890 706 ns/op 251 B/op 3 allocs/op
BenchmarkConcurrentCASMutexLock-4 1677104 714 ns/op 255 B/op 3 allocs/op
BenchmarkConcurrentCASMutexLock-8 1808582 639 ns/op 255 B/op 3 allocs/op
BenchmarkConcurrentCASMutexLock-16 1918422 622 ns/op 255 B/op 3 allocs/op
BenchmarkConcurrent50CASMutexLock 5650274 210 ns/op 96 B/op 1 allocs/op
BenchmarkConcurrent50CASMutexLock-2 1698381 707 ns/op 247 B/op 3 allocs/op
BenchmarkConcurrent50CASMutexLock-4 1697070 707 ns/op 255 B/op 3 allocs/op
BenchmarkConcurrent50CASMutexLock-8 1809859 655 ns/op 255 B/op 3 allocs/op
BenchmarkConcurrent50CASMutexLock-16 1801652 646 ns/op 256 B/op 4 allocs/op
BenchmarkCASMutexTryLock 6593228 182 ns/op 96 B/op 1 allocs/op
BenchmarkCASMutexTryLock-2 7703084 154 ns/op 96 B/op 1 allocs/op
BenchmarkCASMutexTryLock-4 7790750 151 ns/op 96 B/op 1 allocs/op
BenchmarkCASMutexTryLock-8 7756263 152 ns/op 96 B/op 1 allocs/op
BenchmarkCASMutexTryLock-16 7741186 152 ns/op 96 B/op 1 allocs/op
BenchmarkConcurrentCASMutexTryLock 6509828 189 ns/op 96 B/op 1 allocs/op
BenchmarkConcurrentCASMutexTryLock-2 14058355 82.2 ns/op 17 B/op 0 allocs/op
BenchmarkConcurrentCASMutexTryLock-4 16875369 71.4 ns/op 8 B/op 0 allocs/op
BenchmarkConcurrentCASMutexTryLock-8 10985379 109 ns/op 4 B/op 0 allocs/op
BenchmarkConcurrentCASMutexTryLock-16 12405752 104 ns/op 1 B/op 0 allocs/op
BenchmarkConcurrent50CASMutexTryLock 5631146 204 ns/op 96 B/op 1 allocs/op
BenchmarkConcurrent50CASMutexTryLock-2 6518946 182 ns/op 61 B/op 0 allocs/op
BenchmarkConcurrent50CASMutexTryLock-4 6937346 174 ns/op 43 B/op 0 allocs/op
BenchmarkConcurrent50CASMutexTryLock-8 6493034 182 ns/op 39 B/op 0 allocs/op
BenchmarkConcurrent50CASMutexTryLock-16 6177854 202 ns/op 41 B/op 0 allocs/op
PASS
References
- golang/go#6123
- https://github.com/LK4D4/trylock
- https://github.com/OneOfOne/go-utils/tree/master/sync
- https://github.com/lrita/gosync
- https://github.com/google/netstack/blob/master/tmutex/tmutex.go
- https://github.com/subchen/go-trylock
