It is not my cup of tea but one of my colleagues made some
meassurements of the miricom MX-10G and his results were that the
cards is able to capture all 10Gbps traffic (according to RFC2544).
But I think he did not use tcpdump but rather some propietary
application based on the libpcap.
Jan
2010/6/8 Ian Schorr <ian.schorr@xxxxxxxxx>:
> I'm also curious. Although I'm assuming we're in a similar state to
> the capture situation with 1Gig Ethernet 8 or so years, where:
>
> - Few NICs could actually receive and process packets at line speed.
> - A surprisingly large number of NICs refused to enter passive mode,
> and some had problems allocating buffers properly for jumbo frame
> handling.
> - There were few, if any, inline capture solutions that could capture
> bidirectionally, though fiber taps were possible. Copper taps were
> not yet available.
> - For extended captures (more than a few seconds), disk speed is a
> tremendous bottleneck and nowhere near linespeed+full packet capture
> is possible, even with modestly-expensive RAID solutions.
> - Host BUS speeds actually became a problem, either becoming a quick
> bottleneck when others (NIC+disk) were loosened, or exacerbating them.
>
> ...But I have done no testing of my own this time around.
>
> -Ian
>
> On Tue, Jun 8, 2010 at 4:52 PM, Joerg Mayer <jmayer@xxxxxxxxx> wrote:
>> Hello,
>>
>> has anyone experience doing 10G-Ethernet capture with Wireshark? I'm
>> interested in information about the hardware and problems (performance
>> bottlenecks etc).
>>
>> Thanks
>> Joerg
> ___________________________________________________________________________
> Sent via: Wireshark-users mailing list <wireshark-users@xxxxxxxxxxxxx>
> Archives: http://www.wireshark.org/lists/wireshark-users
> Unsubscribe: https://wireshark.org/mailman/options/wireshark-users
> mailto:wireshark-users-request@xxxxxxxxxxxxx?subject=unsubscribe
>
--
Jan Kastil
galloth@xxxxxxxxx
