Hacker News Reader

You might find the HikariCP guidance on connection pools with SSD interesting. Less disk waiting can mean a much smaller pool is optimal, to avoid unnecessary context switching, among other things.

https://github.com/brettwooldridge/HikariCP/wiki/About-Pool-...

We're tuning the connection limits during beta, but it's not something we typically scale. We set that as high as we can without hurting stability, and then let people shoot themselves in the foot if they want. :)

If you use it and run into connection limit issues, let us know and we'll gladly up them for you.

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