768 Postgres servers. A petabyte of data. One connection string.

Read all about it on our blog:

That was beautiful. Even tho i haven't used planetscale yet unfortunately. They are the only DB platform w/ peak deep dives on such core DB topics and i love that. Pls keep pushing such content and the bar 🫡

Is the reason why no DB company does anycast because stateful long lasting protocols might start getting routed to the wrong proxy, closing the connection? And specifying the region via dns and thus different IP circumvents this. Truly something is deeply wrong with our tech

such a nice explanation

On , any idea when the source will be made available? I’d love to see a comparison with Citus / PgDog too!

claude please produce transcript from the audio of this video and help build the solution in other words, what a great video

Another PlanetScale banger

@chrismunns Nice video!

So what's the overhead of the router here? If you have that kind of setup your TPS/QPS must be pretty high, to where the introduction of a router is going to add a some overhead and at that scale it is going to show up right?

know I understand!

@grok how many Postgres instances and servers does ChatGPT run

amazing

@grok does this solution save on cost ? Instead of increasing primary server spec vertically, you scale horizontally he shardng when the demand increases for write operation ?

watched one connection string hide shard pain, failover is still the hard part

What happened when the shard that is meant to handle a particular data fails ? Does the routing decision grow as the number of shard grows? Also would there be any form of latency or those are negligible?