Understanding the Snowflake ID Generation Algorithm and Its Variants (Baidu UIDGenerator, Meituan Leaf‑Snowflake)

Twitter open‑sourced the Snowflake algorithm, a 64‑bit distributed ID generator that divides the ID into four parts: 1 unused bit, 41 bits for a millisecond timestamp (covering 69 years), 10 bits for data‑center and machine identifiers (up to 32 data‑centers and 32 machines each), and 12 bits for a per‑millisecond sequence (up to 4096 IDs).

1 bit: unused, set to 0.

41 bits: millisecond timestamp.

10 bits: 5 bits for data‑center ID, 5 bits for machine ID.

12 bits: sequence number within the same millisecond.

Advantages :

IDs are roughly time‑ordered because the timestamp occupies the high bits.

No reliance on external databases; high performance and stability.

Bit allocation can be customized to fit business needs.

Disadvantages :

Strong dependence on the system clock; clock rollback can cause duplicate IDs or service downtime.

Baidu UIDGenerator (based on an optimized Snowflake): uses future timestamps and a double‑buffer strategy to handle clock rollback and improve generation performance, while integrating MySQL for ID segment allocation.

Advantages : solves clock‑rollback and performance issues.

Disadvantages : depends on a MySQL database.

Meituan Leaf‑Snowflake :

Named “Leaf” after Leibniz’s quote that no two leaves are identical, emphasizing uniqueness.

IDs are obtained via a proxy service that fetches a batch (segment) from the database; a double‑buffer caches the next segment when 10% of the current one is used, ensuring continuous issuance even if the DB temporarily fails.

Linear scalability; performance meets most business scenarios.

64‑bit IDs are monotonically increasing, suitable as primary keys.

High fault tolerance: cached segments allow operation during short DB outages.

Configurable max_id simplifies migration from legacy ID schemes.

Network jitter does not affect segment updates.

Disadvantages :

ID pattern is predictable, potentially leaking issuance volume and reducing security.

How to Choose : For most internal systems, the original Snowflake algorithm is sufficient; if higher security and reliability are required, consider Baidu's UIDGenerator or Meituan's Leaf‑Snowflake.