Questions about hard disks passthrough

[aur3l14no]

My setup is OMV & Other VMs on top of PVE. After some searching and digging I'm still confused that

1. In order to have good performance inside OMV (without using my CPU too much) with regard to accessing hard disks, I should do hard disks pass through. Or is virt-io already good enough?
2. Are there different kinds of passthrough? According to this article (https://kb.vmware.com/s/article/2009226) there seem to be at least two kinds - Physical compatibility RDMs and Virtual compatibility RDMs. I'm wondering what they correspond to in PVE and if both exists in PVE, which should I use.
3. If I want a physical passthrough, I must pass through the whole sata controller or lsi card instead of just some disks, is that correct?

Thanks a lot.

 

[t.lamprecht]

1. In order to have good performance inside OMV (without using my CPU too much) with regard to accessing hard disks, I should do hard disks pass through. Or is virt-io already good enough?

VirtIO-SCSI is quite performant, you should not loose a too relevant performance percentage using it.

If I want a physical passthrough, I must pass through the whole sata controller or lsi card instead of just some disks, is that correct?

You can pass through single disks (block devices), but that has to be done over the command line, using the /dev/... path.

 

[aur3l14no]

Thank you @t.lamprecht for your reply.

I would like to check my understanding. So basically there are three tiers of disks passthroughs:
1) Non-passthrough (VirtIO-SCSI)
2) Single disks passthrough (I heard this is a LUN passthrough instead of physical passthrough which causes guest OS unable to read SMART info)
3) Passthrough entire hba card (guest OS able to read SMART info now)
Also performances increase slightly from 1 -> 2 -> 3.

Is that correct?

Kind regards.