Johannes Heintz

The most straight forward way is to make 3 filecaches. You can merge the ropfetch nodes in one stream so then they will cook in parallel.

Start up time is the time to launch a new Houdini instances. You could try to improve that by not loading too many hdas and make it not pull a lot of stuff from servers or slow drives. Or you could use top services. Once started it will reuse the launched Houdini instance.
https://www.sidefx.com/docs/houdini/tops/services.html [www.sidefx.com]

Should this still be working in newer versions of Houdini? Somehow, it's not working for me.