Global Unique ID

🍀: Previously Implemented in Project

Approaches

• UUID
• Database (SQL)
• Redis 🍀
• Snowflake
• Baidu
  • UidGeneratorDefaultUidGenerator
  • CachedUidGenerator
• Meituan Leaf
  • Leaf-segment: Database
  • Leaf-snowflake

Redis

• Single Thread
• 通過 INCR(), INCRBY() ATOMIC 操作實作 GUID
• 產生的資料是有序的，對排序業務有利

Design Considerations

First Thought

• Might be to use a primary key with the auto_increment attribute in a traditional database.
• Bad Idea:
  • auto_increment does not work in a distributed environment
  • a single database server is not large enough
  • generating unique IDs across multiple databases with minimal delay is challenging

Questions to Ask

• What are the characteristics of unique IDs?
• For each new record, does ID increment by 1?
• Do IDs only contain numerical values?
• What is the ID length requirement?
• What is the scale of the system?

Use Cases

Multi-master Replication

Cons:

• Hard to scale with multiple data centers
• IDs do not go up with time across multiple servers
• Does not scale well when a server is added or removed

UUID

Has a very low probability of getting collision.

Ticket Server

Reference: Ticket Servers: Distributed Unique Primary Keys on the Cheap

Cons:

• Single Point of Failure (SPOF)

Twitter Snowflake Approach

• Divide and conquer

Additional Talking Points

• Clock synchronization (Network Time Protocol)
• Section length tuning
• High availability