I've got three machines active at the moment, a WS and two render nodes.
My solution is to just have a bunch of NVMEs on the WS (got an expansion card because I typically opt for 2TB drives saying "this will surely be enough", it is not) and then if any of the render nodes needs anything it'll take it from the shared drive on the WS.
Any system resources that takes away from the WS is ... Minuscule or I just can't notice it.
If I had multiple WS computer and multiple users then that wouldn't work as well anymore though...
In your case though I'd agree with hurrycat, just buy NVMEs and have everything on local storage. Less faffing around with network setups and coughing up extra money for 10GB networks and what not.
Some sort of a backup for the files could be nice though - you can try doing a RAID or maybe even just saving everything somewhere externally once in a while. Or, and this isn't an advertisement, you can also use something like Backblaze for automatic cloud backups. You can set it to only backup certain drives (for example drives where you have your library) if privacy is a concern.
Not sure how the other options are but Backblaze seems rather affordable to me with their 8$ plans (or is it 10$) and basically unlimited storage. Do double check it on their / their competitors website though.
