@sborden76 @postgresconf Awesome!

Heya! Thanks for reading. I don't have specific numbers, mostly because it's meant to be a high level doc (mentioned in the post). However, these are mostly references from my personal experience from managing 100+ PG clusters, where each cluster manages 10-12 k commits/s at peak with 15-30TB database size at peak, if that helps :) I must say, i totally forgot about connections when writing that post 😂. Perhaps in the next one. I'd be curious to learn about any anecdotes from your experience too 🙏🏽

AGI will be achieved when it can root cause a segfault

RT @tines_hq: We’re thrilled to announce our Series C fundraise and company valuation of $1.125 billion. ✨ WATCH: CEO @eoinhinchy shares w…

@thdxr PlanetScale is legit cracked

@saikatc Stoked for you! All the best

Using a proxy table for counter helps where the only updates that are done are for counters. Something like this (from an internal tech plan) Load testing showed better results, also ofc because reads are extremely fast now, while writes do take a bit of an hit. Be interesting to do a load test and compare the results between updating the single row vs spread across 100 rows/slots.

@rony358 Very cool! Thanks for sharing, I have bookmarked it a few times but then keep missing it :).

RT @rony358: @shayonj We operate a large fleet, although we are not the largest. Architecture described here Sizes…

@isamlambert That has been my sense too. Esp when it comes to QPS. Good examples of larger companies doing sharded setups fwiw.

@arundsharma @filipecrosk @zerodhaonline @RobinhoodApp Very interesting , thanks for sharing

Really cool and impactful work from the folks at @paradedb on how they moved their Tantivy-powered search from the file system to Postgres block storage. Really clear explanation of the challenges and solutions 🚀 (must read)

RT @sunbains: This makes sense from a practical perspective. People obsess over single digit latency like 0-60 mph car acceleration when…

RT @ericfisher: Absolutely perfect flakes today ❄️ #Dendrites

@_mfridman Very nice indeed. A small footprint that takes in flags and falls back to env vars (a feature req if I may) is what I usually need 99% of the time :).

@eatonphil Just these two for now. I found the basics and guides to ingredient sections extremely useful. Especially getting the “science” right of what goes well with what, how to use existing ingredients for the resources required, building good mental models and so on.

I have been recently looking into LSM trees as I think it solves certain use cases very well. Stumbled upon this thread in Postgres ML - I would _love_ a PG compatible (fully) LSM solution if possible. And yes I did try to bake in rocksdb into PG, it "worked" but also failed hilariously :D.