Hi,
the CPU utilization is negligible. top shows less than 0.4 % CPU utilization while receiving on each of the six cores.
We are, by the way, not using any virtual machine so far. We're running both, the PFs and VFs on the same operating system on physical hardware with six
| Intel(R) Xeon(R) CPU | X5650 @ 2.67GHz |
and 6 GB RAM.
The used packet generator is indeed blasting out as many packets as possible. This is why we can only decrease the sending speed by increasing the packet size (see the table before). However this is no problem for ixgbe when no VFs are being used. I'll check out the interrupt counters to investigate this.
I'll update this post as soon I have further information.
Have a nice weekend,
Richard
PS: We're counting the packets on the receiver's side using the information provided by /proc/net/dev.