I checked different hosting solutions, and the best server is - 192 cores / 384 threads - 1.5TB Ram - 4x10TB PCI4 NVME Storage in Raid-0
This is literally a behemoth. Just rent it. Put all your DB, workers, servers, mailers, on this server.
While that is stupid, feel free to use it as a CI tool as well! Running 384 docker containers in parallel to check all your puppeteer/playwright tests
Bonus point, although grey area - run a monero or similar miner in a low priority to cashback 10% of the server cost.
Now serious question - why would you need microservices, cloud, kubernetes, lambdas and other fun things at all if you can get it working as a monolith just for $2K (ok, let it be $5K with extras) ???
Fault tolerance.
Doing something like this—add fault tolerance mechanism(s)—and your potential attack surface is exponentially reduced compared to “typical” microservice architectures.
There are also potential performance gains because there’s a lot less network hopping around.
I’ve done some big data processing using a single big box strategy in the past and it worked very well—as the saying goes, “distribute when you must.”