Choosing the Right Distributed ID Strategy for Microservices

1. JDK built‑in UUID

Java’s UUID generator creates globally unique identifiers by combining the server’s MAC address, current timestamp, and a random component.

Advantages: simple to generate, good performance, globally unique, works well during data migration or system merges.

Disadvantages: stored as strings, low readability, large storage footprint, slower queries, and higher network overhead in microservice environments.

2. Database auto‑increment ID

A dedicated database instance can generate IDs using an auto_increment column; a record is inserted to obtain the primary key.

When traffic spikes, the database becomes a bottleneck, making this approach risky for distributed services.

Advantages: IDs are strictly ordered and easy to implement.

Disadvantages: requires a separate database instance, high cost, limited performance, and potential security concerns; clustering can mitigate but adds complexity.

3. Segment mode

IDs are allocated in batches from the database, with the maximum value stored and the current range kept in memory, often using Redis.

Advantages: reduces database load and improves performance.

Disadvantages: local generation can cause single‑point failures and non‑continuous IDs after service restarts.

4. Redis generation

Redis can generate global IDs using its atomic INCR/INCRBY commands, leveraging its single‑threaded nature.

Advantages: no database dependency, high performance, numeric IDs are naturally ordered, and Redis clusters mitigate single‑point failures.

Disadvantages: adds Redis as a third‑party component and requires additional coding and configuration effort.

5. Snowflake algorithm

Twitter’s Snowflake generates a 64‑bit ID composed of a timestamp (41 bits), a machine identifier (10 bits), and a sequence number (12 bits), yielding roughly 4096 IDs per millisecond per node.

Advantages: simple, fast, time‑ordered IDs, flexible bit allocation, no external dependencies.

Disadvantages: relies on synchronized clocks; clock rollback can cause duplicate IDs, and managing worker IDs across nodes can be cumbersome.

6. Baidu UidGenerator

UidGenerator is a Java implementation based on Snowflake, packaged as a component that supports custom worker‑ID bits and initialization strategies, suitable for containerized environments.

Advantages: globally unique, high availability, high performance, mitigates clock rollback issues.

Disadvantages: includes a built‑in WorkerID allocator that depends on a database during startup; custom implementations can remove this dependency.

7. Meituan Leaf

Meituan’s Leaf component builds on Snowflake with two optimizations: a segment‑based database approach that batches allocations via a proxy server, and a Snowflake variant that rejects ID requests on clock rollback and raises alerts.

Advantages: globally unique, high availability, high performance, uses Zookeeper to handle clock rollback with weak Zookeeper dependency.

Disadvantages: depends on Zookeeper.

8. Zookeeper unique ID

Zookeeper can generate 32‑bit or 64‑bit sequence numbers using node information, which clients can use as unique identifiers.

Advantages: simple implementation.

Disadvantages: requires Zookeeper, involves multiple API calls, and may need distributed locks under high contention, limiting performance.

Overall, ID generation solutions fall into two categories: centralized approaches (e.g., databases, Redis) that produce shorter, incrementing IDs but can become bottlenecks, and decentralized approaches (e.g., UUID) that are simple and highly scalable but result in longer identifiers without natural ordering.