Four million TPS | Postgres.FM 104 |

[ 🇬🇧_🇺🇸 Check out the subtitles - we now edit them, ChatGPT+manually! You can also try RUclips's auto-translation of them from English to your language; try it and share it with people interested in Postgres!]
Nikolay talks Michael through a recent experiment to find the current maximum transactions per second single-node Postgres can achieve - why he was looking into it, what bottlenecks occurred along the way, and ideas for follow up experiments.
Here are some links to things they mentioned:
* How many TPS can we get from a single Postgres node? (Article by Nikolay) / how-many-tps-can-we-ge...
* Chat history with Postgres AI bot postgres.ai/ch...
* Our episode on the overhead of pg_stat_statements and pg_stat_kcache postgres.fm/ep...
* PostgreSQL 17 beta 2 is out www.postgresql...
* PostgreSQL and MySQL: Millions of Queries per Second (about the work by Sveta and Alexander) www.percona.co...
* postgresql_cluster github.com/vit...
* Track on CPU events for pg_wait_sampling github.com/pos...
* The year of the Lock Manager’s Revenge (post by Jeremy Schneider) ardentperf.com...
* Pluggable cumulative statistics (Postgres hackers thread started by Michael Paquier) www.postgresql...
~~~
What did you like or not like? What should we discuss next time? Let us know in the comments, or by tweeting us on @postgresfm / postgresfm , @samokhvalov / samokhvalov and @michristofides / michristofides
~~~
Postgres FM is produced by:
- Nikolay Samokhvalov, founder of Postgres.ai postgres.ai/
- Michael Christofides, founder of pgMustard pgmustard.com/
~~~
This is the video version. Check out postgres.fm to subscribe to the audio-only version, to see the transcript, guest profiles, and more.

[@deadok68]

Hi guys, very proud of you and yours creativity.Found this channel through Nickolay's some 5 years old podcasts and got here, almost 1/3 already listened, thanks

[@RU-qv3jl]

It would also be interesting to use different machines and try it out with the different connection pooling options. I imagine that could be interesting too. You would add latency for sure. Sadly I don’t have the credits to try something like that :(

[@davidfetter]

If that unix socket regression is real, it's very likely a bug.
Also, the fact that there's a huge difference between the TCP version and the unix socket version suggests that there are improvements to be had in the listener code.