Hacker News Reader

nikitaon Neon Serverless Postgres is generally available

(Neon ceo)

They never exceed the limit you set in the number of cpus.

In practice the cost a lot less than max number of cpus * time. Think about area under the curve of cpu(t).

nikitaon Neon Serverless Postgres is generally available

There are some limits. Architecture wise storage indefinitely and compute up to the largest bare metal node on Amazon plus as many read replicas as you want.

nikitaon Neon Serverless Postgres is generally available

(Neon ceo) 1. You should try and see the experience 2. It scales to 0. If you have lots of projects aurora costs really add up 3. It supports branches and integrates with vercel. You can have environment for every dev and every pr. 4. We will be on multiple clouds soon 5. We are just getting started

nikitaon Please bring neon db to GCP

We don’t charge for egress. We are going to be on GCP eventually, but azure will come first

nikitaon Architecture decisions in Neon, a serverless Postgres service

One thing that is very important is what you charge for. Software or service. We are only charging for the cloud service and it has major implications in the “company architecture”. When you charge for software like yuga and roach and for the cloud you end up with two companies in one.

Counterexample is snowflake - they got it right !

nikitaon Architecture decisions in Neon, a serverless Postgres service

This is pretty much the first part of the strategy: database is just a url.

We have more cool stuff coming!

nikitaon Architecture decisions in Neon, a serverless Postgres service

The problem that shared nothing systems lose compatibility with the single node system. So planetscale and yugabyte are not 100% compatible with Postgres. They are also very expensive at the low end hence the free tiers are discontinued.

I have a thread about it on twitter: https://x.com/nikitabase/status/1723579854266966399?s=46

nikitaon Architecture decisions in Neon, a serverless Postgres service

Neon is open source and under a good license: Apache 2

nikitaon Architecture decisions in Neon, a serverless Postgres service

This is a very deliberate effort which is venture funded. It started with a deck three years ago and a team of myself, Heikki and Stas.

Since it’s my second company (prev was singlestore) the early funding was relatively easy. Later funding was done based on our traction.

nikitaon Architecture decisions in Neon, a serverless Postgres service

(neon ceo) Yes you will have an S3 fetch latency for downloading a layer file for such data. However we don't page out aggressively so it's a very unlikely event. Most of the time you will be fetching from a locally attached SSD which is superior and more predictable than EBS.

Where it works super well is the long tail of small and rarely used database - it pushes our costs way down.

Another advantage is that if the whole region goes down the S3 copy survives.

Finally if you separate storage and compute you can be serverless and scale to 0.

nikitaon Transforming Postgres into a Fast OLAP Database

We are happy to try for sure. Analytics workloads really benefit from separation of storage and compute!

nikitaon Bring your own S3 to Neon Postgres

I’m CEO of Neon. Happy to answer questions here

nikitaon Are we at peak vector database?

What volumes are we talking about.

There are ways to speed things up dramatically. Index build just became multithreaded (see above).

We have ideas on what to do with ingest.

Also do you interest from S3 ?

nikitaon Are we at peak vector database?

Lol rare collaboration between neon, AWS, and supabase.

But if Postgres wins we all win!

nikitaon Are we at peak vector database?

What is taking the most time at scale? Is this ingest, index build or lookups ?

nikitaon Are we at peak vector database?

(Neon CEO) It’s about to get a lot better too. Pgvector now supports multi-threaded build

https://github.com/pgvector/pgvector/issues/409#issuecomment...

nikitaon Cutting down AWS cost by $150k per year simply by shutting things off

(CEO of Neon)

We routinely see 10x savings when switching from RDS or Aurora. Especially if you start adding dev environments.

nikitaon Benchmarking Postgres Vector Search Approaches: Pgvector vs. Lantern

Does pgvector use parallel index build in this benchmark ?

nikitaon AWS cancels serverless Postgres service that scales to zero

The buffers will be cached in remote storage - so it will just work likely.

In the future we are considering separating the buffer pool from compute and turning off compute doesn't drop the buffer cache. It's a research project but PolarDB did that for AliCloud.

nikitaon AWS cancels serverless Postgres service that scales to zero

(Neon CEO).

It's possible and here is the recipe: https://twitter.com/nikitabase/status/1725394843285766463

nikitaon AWS cancels serverless Postgres service that scales to zero

(CEO of neon.tech here)

Scaling to zero is tricky. In order to make it work you need to

1. Separate storage and compute which aurora has done

2. Have a proxy between the user and an actual VM in which you host Postgres

3. Spin up this VM after authenticating a connection, stand up Postgres inside the VM and run the query

You need to do all that very quickly- ideally in a couple hundred ms. This can only be accomplished with having a warm pool of VM or having extremely fast microVMs like firecracker.

I believe AWS aurora just uses ec2 and that’s why scale to 0 is turned off

nikitaon Neon – Serverless, Fault-Tolerant, Branchable Postgres

Yes, cold starts is a surprisingly stubborn problem that has many layers. We got it down to 100-200ms. Recently as we move compute from containers to VMs we starting to see more missed so we regressed it a bit.

It is theoretically possible to get it to 50ms range.

nikitaon The challenges of supporting foreign key constraints

Also would be useful to add a disclaimer that you work for PlanetScale.

nikitaon The challenges of supporting foreign key constraints

Neon is not a shared nothing architecture but shared storage. Makes it 100% compatible with Postgres that supports FK constraints since 1989

nikitaon Accidental database programming

My understanding is that we can sync any sqlite state to any other sqlite state using custom built replication.

Is this how it works and how does it update all the web components?

Would it work with all the frameworks or a custom framework is needed?

nikitaon Show HN: AI Proxy with Support for OpenAI, Anthropic, LLaMa2, Mistral

I've been thinking that with OpenAI introducing more and more workflow there will be a company that turns opensource model to API + workflow. I'm wondering if Braintrust vision is that?

nikitaon Vitess 18

Our uptime will improve. We have been keeping up with the growth (2500 databases are created per day). And I insisted to be fully transparent on uptime.

There is a ton of work internally that will improve the uptime as we get to GA.

nikitaon Vitess 18

Neon doesn't use EBS volumes. That's why it has an inherent cost advantage over Planetscale.

nikitaon Vitess 18

Neon will stay true to Postgres. Our storage is sharded and bottomless. You can get a ton of TPS scale with large Postgres compute plus read replicas.

Shared nothing architectures are useless for 99% of the market and break compatibility with the single node databases.