8 releases
0.6.2 | Sep 7, 2024 |
---|---|
0.6.1 | Sep 7, 2024 |
0.5.4 | Sep 7, 2024 |
#757 in Algorithms
283 downloads per month
Used in 4 crates
(3 directly)
46KB
575 lines
SULID: Snowflake-inspired Universally Unique Lexicographically Sortable Identifier
SULID is a unique ID generation algorithm that combines the benefits of ULID and Snowflake. It offers a highly efficient, reliable, and lexicographically sortable identifier, which ensures uniqueness across multiple data centers and machines, making it ideal for high-concurrency distributed environments.
Features
- High Concurrency Support: Efficiently generates unique IDs in high-concurrency environments.
- Time Ordered: Retains the time-ordered characteristic of ULID.
- Distributed Uniqueness: Ensures unique IDs across distributed environments by incorporating data center and machine IDs.
- Readability: Produces shorter, human-readable identifiers.
Overview
SULID is based on the ULID (Universally Unique Lexicographically Sortable Identifier) format but incorporates additional bits for a data center ID and machine ID, similar to Snowflake. This design ensures uniqueness in distributed environments and maintains time-ordering characteristics, making it suitable for applications requiring both high concurrency and global uniqueness, such as microservice architectures, log tracking, and order generation.
ID Format
SULIDs have a unique structure comprising the following parts, adding up to a 128-bit identifier:
Version1
- Timestamp: 48 bits, representing the epoch time in milliseconds.
- Random Number: 70 bits of randomness to ensure uniqueness within the same millisecond.
- Data Center ID: 5 bits, identifying the data center.
- Machine ID: 5 bits, identifying the machine within the data center.
Here is a visual breakdown of the SULID format:
| 48-bit Timestamp | 70-bit Random Number | 5-bit Data Center ID | 5-bit Machine ID |
Version2
- Timestamp: 48 bits, representing the epoch time in milliseconds.
- Random Number: 70 bits of randomness to ensure uniqueness within the same millisecond.
- Worker ID: 10 bits, the combination of data_center_id and machine_id.
Here is a visual breakdown of the SULID format:
| 48-bit Timestamp | 70-bit Random Number | 10-bit Worker ID |
Installation
To use SULID, add the following dependencies to your Cargo.toml
file:
[dependencies]
sulid = "0.6"
Usage
Here's how you can use the SulidGenerator
in your project:
use sulid::SulidGenerator;
fn main() {
let generator = SulidGenerator::v1_new(1, 1);
for _ in 0..3 {
let id = generator.generate();
println!("SULID-V1: {}", id);
}
let generator = SulidGenerator::v2_new(1);
for _ in 0..3 {
let id = generator.generate();
println!("SULID-V2: {}", id);
}
}
Example Output
Running the example code generates SULIDs such as:
SULID-V1: 01J75Y6K09Q0VD4D7HCF496K11
SULID-V1: 01J75Y6K09ZGD0ATFGNJ7TWZ11
SULID-V1: 01J75Y6K09DWG2RPNEXNYSRP11
SULID-V2: 01J75Y6K09FVT8QTQVYGSDJ401
SULID-V2: 01J75Y6K0A16WE3XQN92QPP701
SULID-V2: 01J75Y6K0AN12FBKX56YGZF301
Benefits
- High Concurrency Support: The algorithm is designed to generate IDs efficiently in environments with high concurrency, making it suitable for distributed systems.
- Time Ordered: SULID retains the time-ordering characteristic of ULID. This feature is beneficial for logging systems and event sourcing where chronological order is essential.
- Distributed Uniqueness: By incorporating data center and machine IDs similar to Snowflake's approach, SULID ensures IDs are unique across different machines and data centers.
- Readability: Compared to traditional UUIDs, SULID produces shorter and more human-readable identifiers, making them easier to work with in certain scenarios.
- Traceability: The time-ordered nature of SULID makes it easier to trace and debug events in distributed systems.
How It Works
SULID leverages both ULID and Snowflake's strengths:
ULID
: Provides lexicographically sortable identifiers based on a timestamp and randomness.Snowflake
: Adds data center and machine IDs to ensure distributed uniqueness.
By combining these two approaches, SULID generates IDs that are globally unique, time-ordered, and suitable for high-concurrency distributed environments.
License
This project is licensed under the MIT License.
Dependencies
~41–485KB