起因

我们有一个业务场景(高并发),收到的每个任务都必须在很短的时间内完成,然后回报给调用方,因此每个任务都需要设定定时器。显然golang 默认的timer是不能支持这种场景的,因此我在delayqueue的基础上自己实现了一个库gtimer
GitHub地址

timer

用golang实现的定时器,基于delayqueue

实现

实现受到了Java DelayQueue.java的启发
源码地址
DelayQueue.java

依赖的几个结构依次为为
timer -> delayqueue -> priorityqueue -> heap

由于golang的Condition不支持wait一段时间,所以使用golang原生的Timer来替代了Condition在delayqueue中的作用

Installation

Install:

go get -u github.com/vearne/gtimer

Import:

import "github.com/vearne/gtimer"

Quick Start

package main

import (
    "fmt"
    log "github.com/sirupsen/logrus"
    "github.com/vearne/gtimer"
    "math/rand"
    "strconv"
    "sync/atomic"
    "time"
)

const (
    PRODUCER_COUNT = 10
    CONSUMER_COUNT = 10
    TARGET_COUNT   = 1000000
)

var ops int64 = 0

func main() {
    st := gtimer.NewSuperTimer(CONSUMER_COUNT)
    t1 := time.Now()
    for i := 0; i < PRODUCER_COUNT; i++ {
        go push(st, "worker"+strconv.Itoa(i))
    }

    time.Sleep(100 * time.Millisecond)

    for {
        v := atomic.LoadInt64(&ops)
        if v >= TARGET_COUNT {
            st.Stop()
            break
        } else {
            time.Sleep(100 * time.Millisecond)
        }
    }
    t2 := time.Now()
    log.Infof("cost:%v\n", t2.Sub(t1))
}

func DefaultAction(t time.Time, value string) {
    // fmt.Printf("trigger_time:%v, value:%v\n", t, value)
    atomic.AddInt64(&ops, 1)
}

func push(timer *gtimer.SuperTimer, name string) {
    r := rand.New(rand.NewSource(time.Now().UnixNano()))
    for i := 0; i < 1000000; i++ {
        now := time.Now()
        t := now.Add(time.Millisecond * time.Duration(r.Int63n(300)))
        value := fmt.Sprintf("%v:value:%v", name, strconv.Itoa(i))
        // create a delayed task
        item := gtimer.NewDelayedItemFunc(t, value, DefaultAction)
        timer.Add(item)
    }
}

use NewDelayedItemFunc, we can create a task

// triggerTime is time of the task should be execute
func NewDelayedItemFunc(triggerTime time.Time, value string, f func(time.Time, string)) *Item

task struct like

type Item struct {
    value    string // The value of the item; arbitrary.
    priority int64    // The priority of the item in the queue.
    // The index is needed by update and is maintained by the heap.Interface methods.
    index int // The index of the item in the heap.
    // when task is ready, execute OnTrigger function
    OnTrigger func(time.Time, string)
}

Performance

CPU Model NameCPU ProcessorsMemory

Benchmark Test Results

produce goroutines count consume goroutines count qps(per second)
1 1 285714
10 10 90090
10 100 89285
100 100 23255

PS: 如果更多的考虑性能,推荐大家看看这篇文章
golang:一个高性能低精度timer实现
https://blog.csdn.net/siddontang/article/details/23826507

请我喝瓶饮料

微信支付码