Interview with Matthias van de Meent
- Directing PostgreSQL to Performance: PostgreSQL Index improvements over time
- Wednesday, 17:20
- Berlin 1
- Could you briefly introduce yourself?
I’m Matthias van de Meent, a PostgreSQL hacker from the Netherlands. I’ve been hacking on PostgreSQL since September 2020, which initially consisted of contributing bug fixes.
Since October 2021 I’m full-time employed at Neon to improve PostgreSQL and Neon’s storage system for PostgreSQL.
- How do you engage with the PostgreSQL Community?
I mostly communicate on the -hackers mailing list, but lately I’ve also started attending conferences.
- Have you enjoyed previous PostgreSQL Europe conferences, either as an attendee or as a speaker? (PGConf.EU, FOSDEM PGDay, Nordic PGDay, pgDay Paris, PGConf.DE)
As I’ve only become an active member of the PostgreSQL community in the latter half of 2020, I’ve only been able to physically attend a few conferences. This year, I’ve been able to attend Nordic PGDay, pgDay Paris and PGConf.DE.
- What will your talk be about, exactly? Why this topic?
This talk will go into detail about what performance and space optimizations are known to be still possible in the included index access methods of PostgreSQL.
I personally get quite excited about learning, discovering and applying new optimizations. As indexes themselves are an optimization, they are one of my focus areas in PostgreSQL development.
- What is the audience for your talk?
(Aspiring) PostgreSQL hackers, people interested in (some) AM internals, and people who are interested in the origins of PostgreSQL performance improvements in the area of access methods.
- What existing knowledge should the attendee have?
None of these are strictly required, but a basic familiarity is helpful: A basic understanding of the included index access method, data types, data structures, and the concept of blocks/pages in PostgreSQL.
Some knowledge about optimizing code performance is also helpful, but not required.
- What is the one feature in PostgreSQL 15 which you like most?
If I have to choose from the features in PostgreSQL, then I’d choose transactional DDL. This has existed for a long time, but I really like that we have it, because it is such a headache to migrate data without the guarantee that your modifications can be safely reverted if something goes wrong.
Of the features introduced in PG15, I think I like the new
NULLS [ NOT ] DISTINCTfeature for unique constraints best. It adds compatibility with how unique constraints work in other database systems, and allows the user to define how they want to handle NULLs in their constraints, something that I’ve needed before.
- Which other talk at this year’s conference would you like to see?
There are a lot of great talks, but there are two that I’m really stoked about:
Tomas Vondra’s talk on BRIN improvements and new opclasses. He has done great work on the BRIN access method, and I’m excited to hear more about what he expects in the future.
“On the Road to TDE” by Stephen Frost. Stephen has been advocating for TDE for a long while, and I’ve had some discussions with him on the -hackers mailing list about implementation specifics, so I’m quite interested to hear the latest on this topic.
- Which measure, action, feature or activity would—in your eyes—help to accelerate the adoption of PostgreSQL?
An auto-tuner in core PostgreSQL would be amazing. Right now, we provide a default configuration for the whole database cluster, but these default settings are not so great for many systems. If we could provide an auto-configuration tool that suggests settings at the cluster, database and/or relational level, then that could improve performance without users having to go through the settings where they have the chance to turn autovacuum off.