Alfredo,

At a high level, the results with pause behavior disabled is the same.
 However, we see a couple of interesting differences.  First, the
bandwidth reported by pfcount is much higher.  Second, instead of
seeing pause frames at the switch, we now see dropped packets being
reported in pfcount.

To some extent, this is not a surprising result.  The NIC is sending
pause frames because its buffers are full.  So it seems that we need
to figure out why the NIC thinks its buffers are full and pfcount
thinks that the buffers are empty.  Do you have any suggestions about
how to go about debugging this?  Are there places in the driver that
might be useful to instrument?

Another observation, ARP does not seem to work between the two
machines.  To run our tests we've been manually seeding the ARP cache.

=s=

On Wed, Nov 30, 2011 at 7:11 AM, Alfredo Cardigliano
<[email protected]> wrote:
> Scott
> please try following the order
> sudo ethtool -A dna0 autoneg off
> sudo ethtool -A dna0 rx off
> sudo ethtool -A dna0 tx off
>
> Alfredo
>
_______________________________________________
Ntop-misc mailing list
[email protected]
http://listgateway.unipi.it/mailman/listinfo/ntop-misc

Reply via email to