I'm running rc6 on a Mikrotik HEXs.
HEXs has 4 threads (cpu0->cpu3) and hardware offload.
With standard config, downloading a big iso made cpu0 spike to 100%, the other 3 to 60%, for the duration.
After irqbalance configged, same experiment, cpu0 goes to 60%, the other 3 to 30%, aboutish.
- sqm is disabled
- Soft- and Hardware offload are enabled
- VPN Policy Routing for a Mullvad permanent connection (wireguard) is active
- ddns script for inbound wireguard tunnels
- standard statistics
Questions:
- Can I - or should I want to - redistribute cpu0's load more ? And how ?
- This is weird : "grep eth /proc/interrupts" yields only one interface/interrupt, as if I only had one ethernet interface ?? Using "ip a" I get a list of interface candidates .. but none refers to a second ethernetport ? I don't use the SFP cage, nor does grep sfp.. return anything.
- I had the HEXs from my "Mikrotik fan period" I'm now cured of, and have been using a R2S until now. R2S has faster cpu's than HEXs (higher clocked at least), but no hardware offload. I had high hopes for HEXs beating R2S on openwrt ... but it feels about the same. Yes I use Mullvad which prolly bottlenecks the experience, this is a home-production site with little windows for experimenting. I'll keep using HEXs because I can power it from my PoE switch.
What do you think : without Mullvad, should HEXs beat R2S ?
thx