What are your protips regarding running VyOS as VM-guest?

General questions
,
1

What are your protips regarding optimizations for VyOS when being runned as VM-guest?

Something like:

set system option performance 'virtual-guest'

seems obvious but what else?

Would this considered be good or bad aswell?

set system option kernel cpu disable-nmi-watchdog
set system option kernel disable-hpet
set system option kernel disable-mce
set system option kernel disable-softlockup

And what about offloading per interface?

Should all offloading be turned off when runned as VM-guest or are there any offloading that still helps?

And what about firewall rules?

I will be using services local to the VyOS for my clients (NTP, DNS etc).

So in terms of filtering I would be happy with “screening router” aka regular ACL if compared to L2/L3-switch.

Would firewall rules with notrack be sufficient and how should they be written or will flowtables work here at all?

Differences between KVM (QEMU / Proxmox) and VMware or for that matter XCP-NG?

Im mainly thinking if someone perhaps already benchmarked this and can share your results?

2

Interesting post. I run my two VyOS routers as virtual and I honeslty didn’t realise these options were there. That said, I don’t have any performance problems and I can already reach the maximum throughput availble from the network. Still, very interested to see where this topic goes, and thanks for enlightening me to this raft of option.

3

We run all instances as ESX guests, and neither did I.

In ESX, none of the hardware related interface settings do anything. No offloading, no speed or duplex settings, etc.

The performance option calls “tuned-adm” to set a profile, which is available in every distro as far as I can see. I’ve never used it, and we’ve been running RHEL on ESX for more than 10 years.