Am 14.02.21 um 10:57 schrieb Matthias Petermann:
However, I am still a bit surprised... Interrupts are sent in the case of the Dom0, nevertheless, mainly from the hardware in the direction of the kernel(?) If - as in the test with iperf on localhost (loopback) - no physical I/O hardware is involved at all - how can it be that the difference in this test is so clear?Most likely it's interrupt processing under Xen, or something along those lines.You mean something like what is discussed here:http://www.owlnet.rice.edu/~kevinaboos/docs/interrupt-coalescing-presentation.pdf?
Kind regards Matthias