Hi,
If you are referring to using PCI SSDs as cache, the answer is "no, please don't do that".
Write back RAM caching for SSDs does not make much sense, IMO, and it might make highly available devices a bit touchy, causing full sync or mutual not synchronized every time the restart procedure is mishandled. Write-through caching boosts reads, but in highly available mode, it does not make sense, as client reads from all replication partners, so reads should be OK.
Same for SSD caching, which is available only in write-through mode. Sure, you can try using StarWind HA SSD caching, but that cache boosts only reads, which IMO is a waste of SSDs' potential, as reads are mostly OK due to reading from all replication partners. It is much better to put StarWind highly available devices on top of them to have both reads and writes improved.
p.s. Use pass-through or raw device mapping in Proxmox for data volumes. QUEMU, even with write-back cache, might eat some performance.
p.p.s. If it is for lab use only, try installing opencas inside CVM and give this kind of hardware caching a try. This one should be good if configured right. Again, this is ONLY for lab use, as data loss might be possible in case of hardware failures if that volume is misconfigured.
And, last but not least, if you want to squeeze the maximum performance of your system, join our technical preview for NVMe-oF
https://www.starwindsoftware.com/techni ... ew-program.
Good luck with your project.