The MPLS-OPS Archive

Cell Relay Retreat>MPLS-OPS Archive>month:2002-Jul> msg00079



[Date Prev][Date Next][Thread Prev][Thread Next]  
  [Date Index][Thread Index][Author Index][Subject Index]

Re: Jumbo frames - tag swithing mtu

  • From: Rajiv Asati <rajiva@cisco.com>
  • Date: Wed, 24 Jul 2002 00:43:17 -0400
  • Cc: rogerw@nordlink.com, mpls-ops@mplsrc.com
  • Resent-Date: Wed, 24 Jul 2002 10:12:12 -0400
  • To: CEW CEW <cewsg@yahoo.com>
  • X-Sender: rajiva@dingdong.cisco.com

Eng Wee,

Please see inline..

At 10:00 PM 7/23/2002, CEW CEW wrote:
>Hi...
>
>since we are discussing about mtu, i did some test a
>few days back and a little bit puzzled about the
>behavior.
>
>[r1]----[r2P]-----[r3P]-----[r4P]
>
>tag switching is enabled between r2P, r3P and r4P.
>all links are ethernet link.
>
>I then do a ping sweep from r1 with mtu (1480 to 1540)
>pinging
>to r4P loopback address. All the ping packet can
>go through without a problem. Looks like the router
>r2P is fragmenting the packet. I then do a ping with
>DF bit set and the ping stops at mtu 1496, which make
>sense becos 4 bytes of label are added when r2P send
>the packet out. i then put a "tag-switching mtu 1526"
>command on all tag switching interfaces. With DF bit
>set, the ping stops at mtu 1500.
>
>My questions are :
>(1) Do we need tag-swithing mtu command at all ?

Depends. See below.

>(2) what is the significance of the value used
>in the tag-switching mtu command.

This is specific to the MPLS packets.
This command however assumes that the outgoing interface is capable of 
transmitting a packet bigger than the maximum"IP MTU" size. Depending upon 
the platform and the LC, this assumption may or may not be true.

See the difference in the outputs of "sh int <>" and "sh tag int <> de" and 
notice the MTU size.

I would be interested to know the platform, LC and IOS version you had 
tried with.

Rajiv


>Thanks
>Eng Wee
>
>
>--- rogerw@nordlink.com wrote:
> > Hugo, I may be jumping into a comleted thread, but I
> > will.
> >
> > Every label on an MPLS packet adds 32 bits to the
> > length of the packet. I can think of a possible
> > stack of three labels(MPLS VPN, TE Tunnel, and
> > hop-to-hop MPLS) and that would add 96 bits. If you
> > are running Ethernet for instance, you could be over
> > the standard 1500 and switches and routers that are
> > not MPLS-aware would think these are jumbo frames
> > and therefore out of spec, and delete them.
> >
> > I hope that helps
> >
> > Roger Williams
> >
> > ------------------------------
> > This mobile message sent using PocketMail.
> > Sign up for unlimited e-mail at www.PocketMail.com.
> >
> > -------
> > The MPLS-OPS Mailing List
> > Subscribe/Unsubscribe:
> > http://www.mplsrc.com/mplsops.shtml
> > Archive:
>http://www.mplsrc.com/mpls-ops_archive.shtml
>
>
>__________________________________________________
>Do You Yahoo!?
>Yahoo! Health - Feel better, live better
>http://health.yahoo.com
>
>-------
>The MPLS-OPS Mailing List
>Subscribe/Unsubscribe:  http://www.mplsrc.com/mplsops.shtml
>Archive: http://www.mplsrc.com/mpls-ops_archive.shtml

-------
The MPLS-OPS Mailing List
Subscribe/Unsubscribe:  http://www.mplsrc.com/mplsops.shtml
Archive: http://www.mplsrc.com/mpls-ops_archive.shtml