How UUIDs Hurt Database Performance and What You Can Do About It

Read: How UUIDs can degrade database performance.

The most common way to uniquely identify rows in a database is to use a UUID column, but this approach has performance drawbacks.

What is a UUID?

UUID stands for Universally Unique Identifier. Many versions exist; this article focuses on the popular UUIDv4.

Example of a UUIDv4:

Note: The version number (4) appears in the same position of every UUID.

Issue 1 – Insert Performance

When inserting new records, the index associated with the primary key must be updated to maintain optimal query performance. Indexes are built using B+ trees.

Each record insertion requires rebalancing the underlying B+ tree.

For UUIDs, this rebalancing becomes highly inefficient because the inherent randomness makes it harder to keep the tree balanced. As the dataset grows, millions of nodes may need rebalancing, significantly reducing insert performance.

Issue 2 – Higher Storage Consumption

Consider the size difference between an auto‑increment integer key and a UUID:

An auto‑increment integer consumes 32 bits per value, while a UUID consumes 128 bits.

When stored in a human‑readable format, a UUID can consume up to 688 bits.

This results in roughly four times more storage per row, or up to twenty times when using the readable form.

To evaluate the impact, a simulation was run with two tables, each containing one million rows:

Table 1: rows with UUIDs.

Table 2: rows with auto‑increment integers.

Results:

Total table size: the UUID table is about 2.3 × larger than the integer table.

ID field size: a single UUID field requires 9.3 × more storage than an equivalent integer field.

ID column size (excluding other attributes): the UUID column is 3.5 × larger than the integer column.

Conclusion

UUIDs are an excellent way to ensure uniqueness across records. While the performance and storage issues become noticeable at large scales, they do not cause significant degradation for most projects. Nevertheless, understanding these implications is essential for optimal database design.