Serverless PostgreSQL on Replit | Hacker News Reader

Serverless PostgreSQL on Replit

blog.replit.com

36 pointsamasad9 comments

I have been craving a serverless relational database for low traffic workloads (Wordpress-like sites and simple applications that can be serviced freely by serverless functions).

I have tried AWS's Aurora Serverless V1, but it seems AWS are phasing out support for it as it's not available for newer Postgres versions. It appears that Aurora Serverless V2 (which does not spin down to 0 units and will charge you when not in use) is replacing it.

There's Planetscale and Cockroach DB serverless - both appear compelling.

Would be nice if Amazon offered a similar product so I can use Cloudformation for all my infrastructure provisioning needs.

I guess in AWS-land the true Serverless DB is DynamoDB or Athena with S3? But I was also a bit disappointed by AuroraV2. Even with the minimal config it's still quite expensive if you expect "Serverless" but compared to the other RDS offerings it seems good (in relation to connection limit and HA).

Give us a try (i'm the CEO) - bit.io - we're fully serverless Postgres, and most small workloads will fall within our free tier. Happy to answer any questions.

This is great! Do you support Teraform for provisioning and releasing resources?

For anyone interested, I did a small compilation of some serverless relational DBs here: https://news.ycombinator.com/item?id=31538868

Main issue with neon for us right now is that it doesn't offer granular access because it disallows the creation/assignment of users and roles.

"Under the hood, this product is powered by our friends at Neon who have created a lightning-fast serverless database."

Nice win for Neon.

I never heard of either of these, but Neon does look pretty cool. I might try it for some homelab type stuff that doesn't need a database running 24/7.

https://github.com/neondatabase/neon

What's the difference between the serverless PG of Neon vs something like ElephantSQL, which advertises itself as PG as a service? I'm guessing PG as a service is always running where serverless spins up to service the request? For standard CRUD access patterns does the distinction matter?

History really rhymes. This is another step in Replit reproducing the path of Heroku (which is a good thing to say IMO).

A next generation of IaaS with great DevX would be a relief to everyone burnt by AWS‘s complexity overhead (for a great bunch of customer’s needs).

Reply on news.ycombinator.com