Skip to content

High Scalable In-memory task scheduler using Min Heap and less goroutines

License

Notifications You must be signed in to change notification settings

kawasin73/htask

Repository files navigation

htask (min Heap TASK scheduler)

CircleCI

htask.png

High Scalable In-memory task scheduler using Min Heap implemented in Golang.

htask creates only 1 (scheduler) + n (worker) goroutines, NOT creating goroutines for each task.

if workers size == 0 then scheduler create goroutine for each task when timer have expired.

github.com/kawasin73/htask/cron is wrapper of htask.Scheduler, cron implementation with human friendly interface.

Japanese blog -> Goでスケーラブルなスケジューラを書いた

Install

go get github.com/kawasin73/htask

Cron Usage

package main

import (
	"fmt"
	"sync"
	"time"

	"github.com/kawasin73/htask/cron"
)

func main() {
	var wg sync.WaitGroup
	workers := 1
	c := cron.NewCron(&wg, cron.Option{
		Workers: workers,
	})

	task := func() {
		fmt.Println("hello world")
	}

	// executed every 10:11 AM.
	c.Every(1).Day().At(10, 11).Run(task)

	// task will be executed in every 1 minute from now.
	c.Every(1).Minute().Run(task)

	tenSecondsLater := time.Now().Add(10 * time.Second)
	// executed in every 2 seconds started from 10 seconds later.
	cancel, err := c.Every(2).Second().From(tenSecondsLater).Run(task)
	if err != nil {
		// handle error
	}

	// cron can schedule one time task.
	c.Once(tenSecondsLater.Add(time.Minute)).Run(func() {
		// task can be cancelled.
		cancel()
	})

	c.ChangeWorkers(0)

	time.Sleep(3 * time.Second)

	// on shutdown all queued task will be discarded.
	c.Close()
	wg.Wait()
}

Scheduler Usage

package main

import (
	"context"
	"fmt"
	"sync"
	"time"

	"github.com/kawasin73/htask"
)

func main() {
	var wg sync.WaitGroup
	workers := 1
	scheduler := htask.NewScheduler(&wg, workers)

	ctx, _ := context.WithCancel(context.Background())
	scheduler.Set(ctx.Done(), time.Now().Add(time.Second*2), func(t time.Time) {
		fmt.Println("later executed at :", t)
	})
	scheduler.Set(ctx.Done(), time.Now().Add(time.Second), func(t time.Time) {
		fmt.Println("first executed at :", t)
		// it can set to scheduler while executing task.
		scheduler.Set(ctx.Done(), time.Now().Add(time.Millisecond*500), func(t time.Time) {
			fmt.Println("second executed at :", t)
		})
	})

	scheduler.ChangeWorkers(10)

	time.Sleep(3 * time.Second)

	// on shutdown
	scheduler.Close()
	wg.Wait()
}

Interface

Scheduler Interface

  • func NewScheduler(wg *sync.WaitGroup, workers int) *Scheduler
  • func (s *Scheduler) Set(chCancel <-chan struct{}, t time.Time, task func(time.Time)) error
  • func (s *Scheduler) ChangeWorkers(workers int) error
  • func (s *Scheduler) Close() error

Notes

  • min heap have no limit size.
  • when scheduler is closed, all pending tasks will be discarded.

Benchmarking

# using benchstat
go get -u golang.org/x/perf/cmd/benchstat

# benchmark Scheduler.Set()
go test -bench=. -count=10 > bench.txt && benchstat bench.txt

# benchmark latency
go run cmd/latency/main.go -interval=1000000 -n 10000 -worker=0

benchmark result

$ go run cmd/latency/main.go -interval=1000 -n 1000000 -worker=0
set 1000000 tasks in 2.995069909s. interval = 1µs, total=1s, workers=0
all task have executed in 1.187109126s.
task executed latency : mean=89.401961ms, min=825.441µs, max=187.529808ms
executed min index=0, max index=995060

$ go run cmd/latency/main.go -interval=100000 -n 100000 -worker=0
set 100000 tasks in 292.649906ms. interval = 100µs, total=10s, workers=0
all task have executed in 9.99992153s.
task executed latency : mean=27.207µs, min=12.938µs, max=1.195354ms
executed min index=74683, max index=0

$ go run cmd/latency/main.go -interval=1000000 -n 10000 -worker=0
set 10000 tasks in 28.64034ms. interval = 1ms, total=10s, workers=0
all task have executed in 9.999248894s.
task executed latency : mean=270.313µs, min=30.631µs, max=5.821627ms
executed min index=9118, max index=6563

LICENSE

MIT

About

High Scalable In-memory task scheduler using Min Heap and less goroutines

Topics

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages