You are right. I double check the traces.
1, 4 byte is checksum.
2, The pcap traces are actually broadcasted to another end, so, there
is no flow table. But if I ping first, the end host can authenticate
the dst mac, therefore a route can be set up for later coming pcap
traces.
These are not big issues. Thanks for your reply.

On Wed, Jan 27, 2010 at 1:54 PM, Dan Talayco <[email protected]> wrote:
> Hi Guanyao--
>
> For (1), I expect that the CRC is counted in the stats on the
> switch, but not by pcap.  It is a 4 byte "checksum" (actually,
> cyclical redundancy check) attached at the L2 layer.  This is
> a frequent problem with counting bytes on Ethernet switches.
>
> I'm not sure about (2).  Flows expire and it can sometimes be
> hard to know what's there when a packet hits the switch.  It
> could also be something particular about your hardware setup.
> What hardware are you using?
>
> -Dan
>
> On 1/26/10 1:09 PM, Guanyao Huang wrote:
>>
>> Hi
>> I am running on my own nox module on a topology of 5 switches. I find
>> two phenomena:
>>
>> 1, The byte number of every packet read from statistics is always 4
>> bytes larger than the real size.
>> If pcap shows the packet is of size N,  ./dpctl dump-flows will show a
>> size of N+4. So, if we query statistics, it is also N+4. I want to
>> know why there is a difference of 4?
>>
>> 2, I found that sometimes packets can also get redirected without an
>> entry in flow table, or, the switch has a bug to show the correct flow
>> table.
>> When I tcpreplay from one host to another one, packets get redirected
>> (tcpdump shows so), but  ./dpctl dump-flows shows no entry. Even if
>> the flow is so short to get counted as 0 size (first packet get lost),
>> an entry should still exist.
>> However, if I ping between these two hosts and then tcpreplay, the
>> entries will show correctly.
>> Is there any possible reason why flows get redirected but an entry is
>> not there?
>>
>> Many thanks.
>> _______________________________________________
>> openflow-discuss mailing list
>> [email protected]
>> https://mailman.stanford.edu/mailman/listinfo/openflow-discuss
>

_______________________________________________
nox-dev mailing list
[email protected]
http://noxrepo.org/mailman/listinfo/nox-dev_noxrepo.org

Reply via email to