Skip to content

high-performance common weighted algorithm library

License

Notifications You must be signed in to change notification settings

Danceiny/weighted

 
 

Repository files navigation

License GoDoc travis coveralls Go Report Card

What's New: Update weight dynamically, for exploring better servers at the client end

func ExampleSW_NextWithCallback() {
    // positive: weight more, cost less
    var doFuncPositive = func(input string) int {
        index, _ := strconv.Atoi(input[len(input)-1:])
        cost = 10 - index
        // mock slow operations
        time.Sleep(time.Duration(cost) * time.Microsecond)
        return cost
    }

    w := &SW{}
    w.Add("server5", 100)
    w.Add("server2", 100)
    w.Add("server3", 100)
    results := make(map[string]int)
    for i := 0; i < 9999; i++ {
        so, f := w.NextWithCallback()
        s, _ := so.(string)
        callback := doFuncNegative(s)
        f(1000 / callback) // update weight
    }
}

rust version: weighted-rs

Package weighted implements the smooth weighted round-robin balancing algorithm. This algorithm is implemented in Nginx: https://github.com/phusion/nginx/commit/27e94984486058d73157038f7950a0a36ecc6e35.

Notice: The weighted is NOT goroutine-safe so you MUST use the synchronization primitive to protect it (the Next method) in concurrent cases.

Algorithm is as follows: on each peer selection we increase current_weight of each eligible peer by its weight, select peer with greatest current_weight and reduce its current_weight by total number of weight points distributed among peers.

In case of { 5, 1, 1 } weights this gives the following sequence of current_weight's: (a, a, b, a, c, a, a)

This is an example to use it:

package main

import "fmt"

func ExampleW1_Next() {
	w := &SW{}
	w.Add("a", 5)
	w.Add("b", 2)
	w.Add("c", 3)

	for i := 0; i < 10; i++ {
		fmt.Printf("%s ", w.Next())
	}
}

And this lib has provides another weighted round robin algorithm. This algorithm is used in LVS. It has better performance but it is not so more smooth than the first algorithm, so you can select one algorithm according to your case. It is used like the first:

package main

import "fmt"

func ExampleW2_Next() {
	w := &W2{}
	w.Add("a", 5)
	w.Add("b", 2)
	w.Add("c", 3)

	for i := 0; i < 10; i++ {
		fmt.Printf("%s ", w.Next())
	}
}

About

high-performance common weighted algorithm library

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Go 93.9%
  • Makefile 6.1%