how is divided the data across a cluster?

I know that a 3-node cluster is 3 copies of the whole data but what about a 15-node cluster?
Is it 5 x 3 copies or… ?

The only replication built into Event Store is consistent non-sharded, so a 15 node cluster will have 15 copies of the same data, and a quorum of 8. It will likely not perform very well as writes will have to wait for 7 acknowledgements before acknowledging to a client.

so basically, Event Store can only scale vertically by using more beefy nodes?

Its rather trivial to shard across multiple replica sets. Do you have
a use case?

use case, let’s say use EventStore as main source of truth of all events at Facebook, how do I shard?

If you are at facebook let's chat :wink:

If the max Event Store cluster size is, say 9 ultra-beefy nodes, it’s still a lot of manual config to manually shard for growing parts of the site no ?

That is replication not partitioning.

just a blog post would help to explain some strategies to grow past one cluster…

Pas un blogpost, cela depend de vos besoins. Je rigole etes-vous Facebook?

C'est pas trop difficile a 16 partitions. envoie moi un email private.