the early days will be messy. long term I think it’s viable. self healing cattle is already the norm. why bother humans with alarms and graphs, only to bumble their way through a runbook?
Discussion
you have not seen FoundationDB
not in practice, but I am aware of it. why?
it is extremely hard to break and just works
automatic self-healing and management
it will tell you where the bottleneck is so you can automatically scale it up
setting it up involves copying a single fdb.cluster file and starting it
but is it web scale?
/s
if by web scale you mean Apple CloudKit and Snowflake scale, yes, actually
it in fact does not write your data to /dev/null
it is a KV store, which means you can build a lot, but also means you need to build your own architecture
perfect for nostr use cases
it struggles a bit with large values, but you can easily fix that by using SeaweedFS (some-maintenance-required FS that allows storing a lot of small files like 100KB)