PGConf.Russia 2025

This talk explores the challenges we encountered — and solved — while migrating massive databases from Oracle to PostgreSQL.

One of the most complex aspects of this process was rebuilding foreign keys (FKs). To overcome these challenges, we had to dive deep into the internal workings of FK creation and validation.

Key topics include:

Does ALTER TABLE ... ADD CONSTRAINT ... FOREIGN KEY have an execution plan?

Can an ordinary user influence this process?

What locks are applied during FK creation and validation?

Also, we’ll introduce a first-principles method for investigating performance issues, applying it to a real-world FK creation bottleneck.

This talk will explore different approaches to storing files in PostgreSQL, including:

• Simple table-based storage

• Large objects with pg_largeobject

• pgpro_sfile – a large object (pgpro_bfile) storage solution

In any business-critical system, there comes a need to modify table structures — whether it's adding new columns with expressions based on other columns or converting a non-partitioned table into a partitioned one.

Using standard PostgreSQL tools for large table reorganization often results in significant downtime, as tables become unavailable for writes — and sometimes even for reads — during the process.

In this talk, I’ll introduce dbms_redefinition, a custom-built extension that brings functionality similar to Oracle’s DBMS_REDEFINITION package to PostgreSQL. This solution minimizes downtime to near zero when applying schema changes in production. 

Unlike pg_repack, which does not support structural modifications, dbms_redefinition is useful for all PostgreSQL users, even beyond Oracle migrations.

I’ll share a real-world use case, discuss future development plans, and explore new features in the pipeline.

In PostgreSQL performance testing, benchmarks measure query execution time (latency). To get more reliable results, queries are executed repeatedly, generating a dataset of latency values. Performance is often assessed using standard metrics like the median or mean, but we propose a more advanced approach.

In practice, latency distributions are often multimodal, consisting of multiple underlying distributions with distinct characteristics. In such cases, traditional statistical methods are insufficient, requiring a more detailed analysis of the dataset’s structure.

Our work presents a tool that automatically performs statistical analysis of benchmark results, accounting for dataset-specific features. It detects multimodality, identifies the number and boundaries of dominant modes, and determines key distribution parameters—providing deeper insights into PostgreSQL performance variations.