-
Split 10 bits of Nodes into: 5 bits for IDCs(max to 32 IDCs), 7 bits for nodes(max to 128 nodes)
An IDC can be power down, so better use some IDCs than setting too many nodes into one. -
Decrease 12 bits of Step/Sequence(max to 4096) to 8 bit (max to 256 in 1 ms)
As we tested with following code, it actually only need about 3 step\sequence numbers in 1 ms.
for i := 0; i < 10; i++ {
fmt.Printf("ID : %s\n", node.Generate().Base2())
}In real case, your service gonna do a lot of things after getting an ID. So, 256 is more than enough for a node.
- Increase 41 bits of Time Bits(max to 69 years) to 43 bits(max to 278 years)
snowflake is a Go package that provides
- A very simple Twitter snowflake generator.
- Methods to parse existing snowflake IDs.
- Methods to convert a snowflake ID into several other data types and back.
- JSON Marshal/Unmarshal functions to easily use snowflake IDs within a JSON API.
- Monotonic Clock calculations protect from clock drift.
For help with this package or general Go discussion, please join the Discord Gophers chat server.
This package should be considered stable and completed. Any additions in the future will strongly avoid API changes to existing functions.
By default, the ID format follows the original Twitter snowflake format.
- The ID as a whole is a 63 bit integer stored in an int64
- 43 bits are used to store a timestamp with millisecond precision, using a custom epoch.
- 5 bits are used to store a idc id - a range from 0 through 31.
- 7 bits are used to store a node id - a range from 0 through 127.
- 8 bits are used to store a sequence number - a range from 0 through 255.
You can alter the number of bits used for the idc id, node id and step number (sequence)
by setting the snowflake.IDCBits snowflake.NodeBits and snowflake.StepBits values.
For now, there is a maximum of 20 bits available that can be shared between these two
values. You can change that.
By default this package uses the Twitter Epoch of 1288834974657 or Nov 04 2010 01:42:54.
I updated it to 1612562862000 or 2021-02-06 06:07:42 in milliseconds.
You can set your own epoch value by setting snowflake.Epoch to a time in milliseconds
to use as the epoch.
When setting custom epoch or bit values you need to set them prior to calling any functions on the snowflake package, including NewNode(). Otherwise the custom values you set will not be applied correctly.
Each time you generate an ID, it works, like this.
- A timestamp with millisecond precision is stored using 43 bits of the ID.
- Then the IDC ID is added in subsequent bits.
- Then the Node ID is added in subsequent bits.
- Then the Sequence Number is added, starting at 0 and incrementing for each ID generated in the same millisecond. If you generate enough IDs in the same millisecond that the sequence would roll over or overfill then the generate function will pause until the next millisecond.
The default Twitter format shown below.
+--------------------------------------------------------------------------+
| 1 Bit Unused | 41 Bit Timestamp | 10 Bit NodeID | 12 Bit Sequence ID |
+--------------------------------------------------------------------------+
My Updated format shown below.
+---------------------------------------------------------------------------------------+
| 1 Bit Unused | 43 Bit Timestamp | 5 Bit IDCID | 7 Bit NodeID | 8 Bit Sequence ID |
+---------------------------------------------------------------------------------------+
Using the default settings, this allows for 256 unique IDs to be generated every millisecond, per Node.
This assumes you already have a working Go environment, if not please see this page first.
go get github.com/butaixianran/snowflakeImport the package into your project then construct a new snowflake Node using a unique node number. The default settings permit a node number range from 0 to 1023. If you have set a custom NodeBits value, you will need to calculate what your node number range will be. With the node object call the Generate() method to generate and return a unique snowflake ID.
Keep in mind that each node you create must have a unique node number, even across multiple servers. If you do not keep node numbers unique the generator cannot guarantee unique IDs across all nodes.
Example Program:
package main
import (
"fmt"
"github.com/butaixianran/snowflake"
)
func main() {
// Create a new Node with a IDC number of 3 and a Node number of 1
node, err := snowflake.NewNode(3, 1)
if err != nil {
fmt.Println(err)
return
}
// Generate a snowflake ID.
id := node.Generate()
// Print out the ID in a few different ways.
fmt.Printf("Int64 ID: %d\n", id)
fmt.Printf("String ID: %s\n", id)
fmt.Printf("Base2 ID: %s\n", id.Base2())
fmt.Printf("Base64 ID: %s\n", id.Base64())
// Print out the ID's timestamp
fmt.Printf("ID Time : %d\n", id.Time())
// Print out the ID's idc number
fmt.Printf("ID IDC : %d\n", id.IDC())
// Print out the ID's node number
fmt.Printf("ID Node : %d\n", id.Node())
// Print out the ID's sequence number
fmt.Printf("ID Step : %d\n", id.Step())
// Generate and print, all in one.
fmt.Printf("ID : %d\n", node.Generate().Int64())
// Or do it with a loop
for i := 0; i < 10; i++ {
fmt.Printf("ID : %s\n", node.Generate().Base2())
}
}With twitter's default settings, this snowflake generator should be sufficiently fast enough on most systems to generate 4096 unique ID's per millisecond. This is the maximum that the snowflake ID format supports. That is, around 243-244 nanoseconds per operation.
Since the snowflake generator is single threaded the primary limitation will be the maximum speed of a single processor on your system.
To benchmark the generator on your system run the following command inside the snowflake package directory.
go test -run=^$ -bench=.If your curious, check out this commit that shows benchmarks that compare a few different ways of implementing a snowflake generator in Go.