Postgres In 2024 | Scaling Postgres 297

To get the show notes as well as get notified of new episodes, visit:

In this episode of Scaling Postgres, we discuss hopes for Postgres in 2024, whether you need foreign keys, incremental backups, and five ways of doing not exists.

Want to learn more about Postgres performance?

Timestamps:
00:00 - Intro
00:15 - Thoughts on PostgreSQL in 2024
07:39 - Do you really need Foreign Keys?
11:05 - PostgreSQL 17: Support for incremental backups
13:58 - Waiting for Postgres 17: Incremental base backups
14:22 - 5 Ways to implement NOT EXISTS in PostgreSQL
16:33 - Praise, Criticism, and Dialogue
17:28 - Partitioning by ULID
18:31 - work_mem
19:48 - Demand the impossible: rigorous database benchmarking
20:17 - Outro

#postgres #postgresql

[jaimeduncan]

The blog about the Postgres status does not mention write amplification that kills Postgresql in many instances. Does not talk about indirect indexes or the implementation of MVC is like the full PG community has the head under a wrap on the full negation of the status. The UBER article is 8 years old and nothing has been done. In October of the same year (2016), Alvaro Herrera showed a code implementation of indirect indexes showing massive performance gains. We do have the work that is being done by EDB but not even they mention the problem of write amp

[professortrog]

Though ULID is better then UUID's, a smarter way is to use a bigint that incorporates a timestamp or date on the left side.