Google Translate:
I have a PVE8.4 system and installed 64 VMs, each with 4 cores. I want to ensure that each VM has at least 2 cores of basic resource protection under high load on the server. What should I do?
Each VM has just a bunch of Linux processes, so a resource scheduling like you descibed is not possible directly in PVE and may be counter productive. All processes / VMs share the resources so that all will be slower if over committed, yet no one can starve the other. If you want to mess with the linux scheduling, look into renice in order to give single processes a higher nice level, yet this cannot be done via PVE directly.