14 comments

[ 5.3 ms ] story [ 34.3 ms ] thread
this is great to see. thank you for the write up.
> Beyond reliability, we faced other challenges:

> - Usage-based pricing that punished our success, the more users chatted, the more we paid

This is such a strange position on usage-based pricing and seems telling.

Neon doesnt have usage based pricing it has autoscaling, there is a difference.
What exactly is PlanetScale Postgres? Is it plain managed Postgres ala RDS or something more bespoke like Neon? I know PlanetScale is working on a Vitess-like sharded Postgres (Neki?) but I’m guessing that is not yet running in the cloud?
They can't afford database administration? Is it that expensive to hire a DBA?
> We have thousands of users generating thousands of chats daily. Lawyers discussing sensitive cases. Executives planning strategy. Developers working on proprietary code. They trust us because their data is mathematically guaranteed to be private. But that trust evaporates quickly when the service is unavailable.

And 250 dollars/month was considered expensive for the infrastructure handling that? My first impression is that based on the stakes alone, that would warrant a full time dedicated engineer.

Not gonna lie, although I appreciate the comparison and that they shared their experience publicly, this post sounds like half a technical write-up, half an ad for both companies.

We're experimenting with Lakebase now (Databricks' name for Neon).

Initial results are actually pretty cool when using the UI to spin it up. The API leaves some things to be desired (bad error messages that obfuscate the actual error, undocumented rate limiting, ..).

Plus, there's been quite a number of strange bugs we ran into, like tables that don't recreate correctly because of dangling resources.

Overall, I'm pretty excited about the product because it makes life a bit easier, but it's not really 'production quality yet'.

(Alternatively, maybe we're just doing things in a bad way while we're learning to use it, so it could be a PEBKAC type of issue, rather than a Lakebase issue).

(comment deleted)
I realize this wasn't the main reason they switched, but from my perspective, $156/mo and $250/mo for db is essentially the same number:

Way too much for a project without a budget, and approximately zero for a project with a budget.

I thought PlantScale's postgres option was still in early access mode?
TiDB is another option in this space which claims to be even cheaper and auto-scaling than PlanetScale.