ksuid

ksuid is an efficient, comprehensive, battle-tested Go library for generating and parsing a specific kind of globally unique identifier called a KSUID. This library serves as its reference implementation.

Install

go get -u github.com/segmentio/ksuid

What is a KSUID?

KSUID is for K-Sortable Unique IDentifier. It is a kind of globally unique identifier similar to a RFC 4122 UUID, built from the ground-up to be "naturally" sorted by generation timestamp without any special type-aware logic.

In short, running a set of KSUIDs through the UNIX sort command will result in a list ordered by generation time.

Why use KSUIDs?

There are numerous methods for generating unique identifiers, so why KSUID?

Naturally ordered by generation time
Collision-free, coordination-free, dependency-free
Highly portable representations

Even if only one of these properties are important to you, KSUID is a great choice! :) Many projects chose to use KSUIDs just because the text representation is copy-and-paste friendly.

1. Naturally Ordered By Generation Time

Unlike the more ubiquitous UUIDv4, a KSUID contains a timestamp component that allows them to be loosely sorted by generation time. This is not a strong guarantee (an invariant) as it depends on wall clocks, but is still incredibly useful in practice. Both the binary and text representations will sort by creation time without any special sorting logic.

2. Collision-free, Coordination-free, Dependency-free

While RFC 4122 UUIDv1s do include a time component, there aren't enough bytes of randomness to provide strong protection against collisions (duplicates). With such a low amount of entropy, it is feasible for a malicious party to guess generated IDs, creating a problem for systems whose security is, implicitly or explicitly, sensitive to an adversary guessing identifiers.

To fit into a 64-bit number space, Snowflake IDs and its derivatives require coordination to avoid collisions, which significantly increases the deployment complexity and operational burden.

A KSUID includes 128 bits of pseudorandom data ("entropy"). This number space is 64 times larger than the 122 bits used by the well-accepted RFC 4122 UUIDv4 standard. The additional timestamp component can be considered "bonus entropy" which further decreases the probability of collisions, to the point of physical infeasibility in any practical implementation.

Highly Portable Representations

The text and binary representations are lexicographically sortable, which allows them to be dropped into systems which do not natively support KSUIDs and retain their time-ordered property.

The text representation is an alphanumeric base62 encoding, so it "fits" anywhere alphanumeric strings are accepted. No delimiters are used, so stringified KSUIDs won't be inadvertently truncated or tokenized when interpreted by software that is designed for human-readable text, a common problem for the text representation of RFC 4122 UUIDs.

How do KSUIDs work?

Binary KSUIDs are 20-bytes: a 32-bit unsigned integer UTC timestamp and a 128-bit randomly generated payload. The timestamp uses big-endian encoding, to support lexicographic sorting. The timestamp epoch is adjusted to March 5th, 2014, providing over 100 years of life. The payload is generated by a cryptographically-strong pseudorandom number generator.

The text representation is always 27 characters, encoded in alphanumeric base62 that will lexicographically sort by timestamp.

High Performance

This library is designed to be used in code paths that are performance critical. Its code has been tuned to eliminate all non-essential overhead. The KSUID type is derived from a fixed-size array, which eliminates the additional reference chasing and allocation involved in a variable-width type.

The API provides an interface for use in code paths which are sensitive to allocation. For example, the Append method can be used to parse the text representation and replace the contents of a KSUID value without additional heap allocation.

All public package level "pure" functions are concurrency-safe, protected by a global mutex. For hot loops that generate a large amount of KSUIDs from a single Goroutine, the Sequence type is provided to elide the potential contention.

By default, out of an abundance of caution, the cryptographically-secure PRNG is used to generate the random bits of a KSUID. This can be relaxed in extremely performance-critical code using the included FastRander type. FastRander uses the standard PRNG with a seed generated by the cryptographically-secure PRNG.

NOTE: While there is no evidence that FastRander will increase the probability of a collision, it shouldn't be used in scenarios where uniqueness is important to security, as there is an increased chance the generated IDs can be predicted by an adversary.

Battle Tested

This code has been used in production at Segment for several years, across a diverse array of projects. Trillions upon trillions of KSUIDs have been generated in some of Segment's most performance-critical, large-scale distributed systems.

Plays Well With Others

Designed to be integrated with other libraries, the KSUID type implements many standard library interfaces, including:

Stringer
database/sql.Scanner and database/sql/driver.Valuer
encoding.BinaryMarshal and encoding.BinaryUnmarshal
encoding.TextMarshal and encoding.TextUnmarshal (encoding/json friendly!)

Command Line Tool

This package comes with a command-line tool ksuid, useful for generating KSUIDs as well as inspecting the internal components of existing KSUIDs. Machine-friendly output is provided for scripting use cases.

Given a Go build environment, it can be installed with the command:

$ go install github.com/segmentio/ksuid/cmd/ksuid

CLI Usage Examples

Generate a KSUID

$ ksuid
0ujsswThIGTUYm2K8FjOOfXtY1K

Generate 4 KSUIDs

$ ksuid -n 4
0ujsszwN8NRY24YaXiTIE2VWDTS
0ujsswThIGTUYm2K8FjOOfXtY1K
0ujssxh0cECutqzMgbtXSGnjorm
0ujsszgFvbiEr7CDgE3z8MAUPFt

Inspect the components of a KSUID

$ ksuid -f inspect 0ujtsYcgvSTl8PAuAdqWYSMnLOv

REPRESENTATION:

  String: 0ujtsYcgvSTl8PAuAdqWYSMnLOv
     Raw: 0669F7EFB5A1CD34B5F99D1154FB6853345C9735

COMPONENTS:

       Time: 2017-10-09 21:00:47 -0700 PDT
  Timestamp: 107608047
    Payload: B5A1CD34B5F99D1154FB6853345C9735

Generate a KSUID and inspect its components

$ ksuid -f inspect

REPRESENTATION:

  String: 0ujzPyRiIAffKhBux4PvQdDqMHY
     Raw: 066A029C73FC1AA3B2446246D6E89FCD909E8FE8

COMPONENTS:

       Time: 2017-10-09 21:46:20 -0700 PDT
  Timestamp: 107610780
    Payload: 73FC1AA3B2446246D6E89FCD909E8FE8

Inspect a KSUID with template formatted inspection output

$ ksuid -f template -t '{{ .Time }}: {{ .Payload }}' 0ujtsYcgvSTl8PAuAdqWYSMnLOv
2017-10-09 21:00:47 -0700 PDT: B5A1CD34B5F99D1154FB6853345C9735

Inspect multiple KSUIDs with template formatted output

$ ksuid -f template -t '{{ .Time }}: {{ .Payload }}' $(ksuid -n 4)
2017-10-09 21:05:37 -0700 PDT: 304102BC687E087CC3A811F21D113CCF
2017-10-09 21:05:37 -0700 PDT: EAF0B240A9BFA55E079D887120D962F0
2017-10-09 21:05:37 -0700 PDT: DF0761769909ABB0C7BB9D66F79FC041
2017-10-09 21:05:37 -0700 PDT: 1A8F0E3D0BDEB84A5FAD702876F46543

Generate KSUIDs and output JSON using template formatting

$ ksuid -f template -t '{ "timestamp": "{{ .Timestamp }}", "payload": "{{ .Payload }}", "ksuid": "{{.String}}"}' -n 4
{ "timestamp": "107611700", "payload": "9850EEEC191BF4FF26F99315CE43B0C8", "ksuid": "0uk1Hbc9dQ9pxyTqJ93IUrfhdGq"}
{ "timestamp": "107611700", "payload": "CC55072555316F45B8CA2D2979D3ED0A", "ksuid": "0uk1HdCJ6hUZKDgcxhpJwUl5ZEI"}
{ "timestamp": "107611700", "payload": "BA1C205D6177F0992D15EE606AE32238", "ksuid": "0uk1HcdvF0p8C20KtTfdRSB9XIm"}
{ "timestamp": "107611700", "payload": "67517BA309EA62AE7991B27BB6F2FCAC", "ksuid": "0uk1Ha7hGJ1Q9Xbnkt0yZgNwg3g"}

License

ksuid source code is available under an MIT License.

OlegGorj/ksuid