We don't apply any artificial limit on the number of connections, or the number of ops/s.

We scale on storage / memory, we try to keep the storage and memory scaling to a defined ratio for optimal performance