>UUID v7 as Database Primary Keys: Performance and Sortability

✓ Last tested: June 2026 · Verified against RFC 9562

1. Field Notes: The UUID v4 Indexing Nightmare

Choosing the correct unique identifier layout is vital for database scalability. Many development teams default to UUID v4 for primary keys because they are easy to generate and mathematically guarantee uniqueness across distributed systems without coordination.

However, at scale, this choice becomes an architectural bottleneck. Traditional UUID v4 identifiers are completely random. When a relational database like PostgreSQL or MySQL attempts to insert a random string into a B-Tree index, it causes constant index page splitting. The database thrashes its disk I/O, writing to random locations instead of appending sequentially.

2. Enter UUID v7: The Best of Both Worlds

A uuid v7 database primary key generator solves this problem by introducing a time-ordered prefix.

UUID v7 begins with a 48-bit Unix timestamp in milliseconds. Because the leading bits represent time, generated UUIDs are naturally sequential. When inserted into a database, the new records append neatly to the end of the index, mimicking the performance of a traditional auto-incrementing integer while maintaining the distributed generation benefits of a UUID.

3. How to Generate UUID v7 Securely

To prevent predictable collision clustering, the remaining 74 bits of a UUID v7 must be generated using a cryptographically secure random number generator (like the browser's crypto.getRandomValues).

By adopting a robust bulk generator tool that correctly implements RFC 9562, you can safely generate thousands of UUID v7 identifiers that will keep your database indexes perfectly optimized, avoiding the dreaded fragmentation penalty.