[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: End System PMTUD behavior question

Hi Jeffrey,
I have attached a picture which shows one of the topologies we use for our PMTUD tests.
In this test case we transmit an Echo Request from REF-Host2 to TAR- Host1 with a packet size of 1500 bytes. REF-Host2 fragments the Echo Request it transmits. TAR-Host1 replies to this Echo Request with an Echo Reply to REF-Host2 with a size of 1500 bytes that is not fragmented. TAR-Router1 sends a Packet Too Big message in response as this Echo Reply is too large to forward onto Network 2.
In all of the cases we have seen TAR-Host1 does fragment future Echo Replies to REF-Host2, however, it does not retransmit any Echo Replies for Echo Requests received prior to receiving the Packet Too Big message from TAR-Router1.
Additionally from the tests we have performed in our lab if TAR-Host1 were to send an Echo Request with a packet size of 1500 bytes TAR- Router1 would send a Packet Too Big message in response. In all cases we have seen TAR-Host1 would not re-transmit this Echo Request and this would be counted as packet loss in the ping command results.
If this does not ideally match your test scenario we'd be happy to work together off-line to replicate your scenario. 

Thanks,
Tom

PNG image



On Jan 21, 2009, at 5:24 PM, Dunn, Jeffrey H. wrote:


Thomas,
We would be happy to see any results you may have concerning application behavior during PMTUD. 

Best Regards,

Jeffrey Dunn
Info Systems Eng., Lead
MITRE Corporation.
(301) 448-6965 (mobile)

-----Original Message-----
From: Thomas Peterson [mailto:thomasp@iol.unh.edu]
Sent: Wednesday, January 21, 2009 3:00 PM
To: Dunn, Jeffrey H.
Cc: Rémi Denis-Courmont; ipv6@ietf.org; Huang, Frank; Sherman, Kurt T.; Liou, Chern; steve_eiserman@uscourts.gov; ipv6-bounces@ietf.org; v6ops@ops.ietf.org ; Grayeli, Parisa 
Subject: Re: End System PMTUD behavior question

Hi Jeffrey,

This is Thomas Peterson from the InterOperability Laboratory at UNH.

We have done extensive testing in this area and would be happy to work
with you off line to examine this scenario.

If you would like we can even set up your test topology in our lab.

Thanks,
Tom

On Jan 21, 2009, at 2:44 PM, Dunn, Jeffrey H. wrote:


Rémi,

I agree with you assertions concerning "ping" versus TCP; however, I
am looking for some concrete documentation or experiences.

Best Regards,

Jeffrey Dunn
Info Systems Eng., Lead
MITRE Corporation.
(301) 448-6965 (mobile)


-----Original Message-----
From: Rémi Denis-Courmont [mailto:rdenis@simphalempin.com]
Sent: Wednesday, January 21, 2009 2:33 PM
To: ipv6@ietf.org
Cc: Dunn, Jeffrey H.; ipv6-bounces@ietf.org; v6ops@ops.ietf.org;
Sherman, Kurt T.; Liou, Chern; steve_eiserman@uscourts.gov; Huang,
Frank; Grayeli, Parisa
Subject: Re: End System PMTUD behavior question

Le mercredi 21 janvier 2009 20:56:23 Dunn, Jeffrey H., vous avez
écrit :

Colleagues,

We have been performing some PMTUD tests and have found that
different
operating systems handle PMTUD differently. Specifically, we found
that the
"ping" application behaves in the following way when the PMTU is
set to
1280 and a 1500 octet ICMPv6 echo request is sent to that routed
path. The
hosts we tested:


I'm afraid ping applications are not representative of what the
operating
system IP stacks do. Besides, there are no standards regarding the
behavior
of "ping" in case of Path MTU problems; it's really up to the
implementor of
each "ping" whether to report an error or to transparently fragment.


The behavior of real transport protocols, such as TCP, DCCP and SCTP
is more
interesting when it comes to evaluating operating systems and their IP 
stacks.

--
Rémi Denis-Courmont
http://www.remlab.net/
--------------------------------------------------------------------
IETF IPv6 working group mailing list
ipv6@ietf.org
Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------