My Coding Cult

Hello bois.

I started building out my products with EdenPlatform on top of AWS. I started with only 4 servers, two datacenters for quorum, modest 8GB machines while having beefy bare metals at local DC.

It started out with 1000$ per month, which is already unbelievable as any decent VPS provider would provide such hardware at laughable ~15$ per box, which would be 60 dollars total. But whatever, as long as I could get EdenPlatform working with multiple datacenters, including AWS I figured its a proof of concept that you can easily connect expensive AWS cloud to other clouds or datacenters.

The cost was starting to run at unbelievable 2000$ per month for four servers. Some AWS fag would scream "visit cost explorer, yoo moost bee doin sumthin wraaang!".

Yes, I was doing something wrong - entrust my businesses with AWS in the first place. I will not bother with cost explorer and complex AWS pricing rules - thank you.

So, I added hetzner bare metal server implementation to Eden platform, took a week. https://github.com/cultleader777/EdenPlatform/commit/6da40c742c118454310d66d8402d20789e566597

Then, once I got it fully working inside Eden platform, it took me a day to migrate all of my infrastructure from AWS to Hetzner machines for which I'm paying laughable 200 dollars per month for all machines.

I migrated:
- Nomad servers
- Consul servers
- Vault servers
- Victoria metrics servers
- Alertmanager servers
- Minio servers
- Postgres servers
- Clickhouse keeper servers - the only one that involved manual action of reconfiguring peers

And I said goodbye to AWS overpriced garbage VMs by running terraform destroy with a perfectly satisfied smiling face.

Mind you, these machines have twice the cores and 8 times more RAM than AWS pile of garbage small machines have. They perform much better and are always snappy, not at the modest RAM limit barelly chugging like they were in AWS.

So, in my case, price reduction for hardware as not just 90%, I got a lot more powerful and cheap machines and I still have all the properties of Eden Patform, cost ratio is more like hundred times more expensive in AWS.

And AWS faggots will start screaming "but your database is not scalable, your queues are not scalable, your storage is not scalable!"

First and foremost, if you use RDS instance in AWS it is not scalable either. If RDS instance breaks down due to traffic, guess who needs to resize it? You. If kafka instance is no longer handling the load, who needs to resize it? You. If ElasticCache instance is too small, who needs to resize it? You. Who needs to setup prometheus exporters to these services? You.

There are very few AWS services that scale out of the box, like S3 and SQS, but cost of running any server in AWS is laughable. And if you use something like DynamoDB, where you're literally raped for every single request, if your traffic bursts you'll never get better price proposition than if you ran bare metal Postgres instance to begin with.

And guess what? Since I have OCaml programming language that generates EdenPlatform code I could easily code OCaml functions that could burst up compute from any cloud at any time with regarding metrics. This is not just some basic autoscaling like kubernetes, I can write OCaml code spin up entire datacenters and terraform code will be generated and backend application scaled to only compute machines. I have all the power in my hands to do any autoscaling strategy I want with EdenPlatform, even querying moon phases during compile time.

In the context of EdenPlatform, clouds like AWS or Google cloud are legacy systems. Temporary stops. You already have managed highly available Postgres with Patroni, you already have S3 (MinIO), you already have Kafka (self hosted NATS), you already have Analytics database (Clickhouse), you already have DNS (Bind), you already have redundant links across datacenters (Wireguard + BGP routing) and everything is set up in highly available fashion, any server can go down and I'm still happily chugging along. Not to mention everything is using rock solid NixOS to have reproducible state on all machines where you can rest assured operating system will be running a long time without any problems.

In 10 years time Cloud will be looked into as the biggest fraud in the history of computing. Maybe AI will be the first and cloud second, both are unbelievable frauds. Startups that serve 10 requests per second spend 100k on AWS while they could be serving the same for under a thousand dollars on normal hosting (if they also adjusted their mindsets to never run garbage of NodeJS and to run Rust programs instead). Today's DevOps practices will be looking like a clown show 10 years from now, when most hosted services will be run in self hosted manner and where compiler like Eden Platform will check 99% of mistakes for you and deployments will work the first time.

Dear AWS, I always hated you. You were overpriced, useless garbage from day one. And people who praise you are mindless imbeciles. Goodbye, and I hope I'll never see you again.

Good day bois