snowflake

中文说明

This is the snowflake algorithm of a GO language implementation.

Support for more flexible generation methods.

goal：

1. In a distributed environment, the sequence number is unique
2. In a multithreading environment,the sequence number is unique
3. Standalone performance is good enough

implement solution：

The generated serial number is a 64-bit number.In order to achieve this goal, this number is made up of the following components:

• datacenter
• machine
• timestamp
• sequence

Datacenter and machine to achieve multi-machine cases, each machine has a unique identifier;

Timestamps ensure orderly growth and avoid repetition in the case of a single machine;

The sequence is guaranteed to increase in unit time;

In addition, the program must ensure that the sequence generation of multiple processes is not repeated;

Sequence number resolution：

The binary structure of the serial number：

0 0000 0000 0000000000000000000000000000000000000000 000000000000000

Group by space：

• The first group，The value is fixed to 0 to avoid data overflow and negative values
• The second group，The length is not fixed.Be required. datacenter
• The third group，The length is not fixed.Be required. machine
• The fourth group，The length of the 38-42.Be required.The value of the current timestamp minus the fixed timestamp
• The fifth group，The length is not fixed.You don't need to fill it out. It's automatic。Sequence growth at the current time

Installation

go get github.com/agclqq/snowflake

usage

package main

import (
	"fmt"
	"github.com/agclqq/snowflake"
)

func main()  {
	sf,err:=snowflake.New(2,2,2,2,snowflake.T38)
	if err!=nil{
		fmt.Println(err)
		return
	}
	id:=sf.GetId()
	fmt.Println(id)
}

best practice

Efficiency

Using the TestSnowFlake_simg_GetId() method in the test case, you can measure the num value per second that you can achieve by adjusting the parameter New() and the num variable in the program to confirm your desired performance.

In this test case, the hardware environment of 'Intel(R) Core(TM) I7-10750H CPU @ 2.60GHz' can produce about 2.4 million serial numbers per second.

Multiple conflict

If your terminal information is not controllable, such as deploying in k8s pod, you can use datacenter and machine to use random numbers to resolve conflicts. When random digits reach a reasonable length, the probability of conflict is also greatly reduced.

In this test case TestSnowFlake_GetRandomI(),the probability of conflict is about 1 in 900,000