[c-nsp] MPLS MTU

Saku Ytti saku+cisco-nsp at ytti.fi
Tue Nov 14 15:20:39 EST 2006


On (2006-11-14 11:07 -0700), Sergio D. wrote:

> hey thanks. I been using MPLS for a while I just started using Cisco
> as I am new to this network, I knew of this short coming. thank god
> for the adjust mss command that didn't exist in Juniper. Anyway, how
> would I check the current MTU for the PHYSICAL interface, so the
> actual ethernet MTU, and how would I change that?

Cisco calculates MTU differently than JNPR, L2 overhead is 'free',
as you probably guessed by looking 'sh int' and seeing 1500 bytes,
which means you can transport 1500 bytes of payload really.
Honestly I think JNPR's way is insane, either you calculate
full L2 overhead, or you ignore l2 overhead, not 'oh, lets
count L2 but not CRC'.

General rule of thumb, 'mpls mtu' is bad for you, just increase
physical mtu. But for reason I still don't understand even after
reading explanation few times, on some FE interfaces in cisco
you can't increase interface mtu, but you can increase 'mpls mtu'
and infact you can send them unfragmented as-is. 

So if you want to support just L3 MPLS VPN, 1508 is enough for you
(mtu 1508 or mpls mtu 1508 both will work just the same)
But if you want eg. EoMPLS with VLAN you need 18 bytes more, 
add 4 bytes for MPLS FRR and you're at 1530, some ATOM stuff
are fatter than EoMPLS, so thats still not safe for all applications
of MPLS.
I'd just run biggest common denominator per link.

> Thanks again,
> 
> On 11/14/06, Lasher, Donn <DLasher at newedgenetworks.com> wrote:
> >
> > Welcome to mpls. :)
> >
> >
> > You have to make sure that ALL Devices, L2, are able to support the
> > larger MTU. On older catalyst, you set the MTU on the port, on newer on
> > the whole switch.
> >
> > Then you can raise the MPLS MTU on the router interfaces. I'd recommend
> > as big as the L2 devices allow, to avoid any future issues.
> >
> > Good luck..
> >
> >
> >
> > -----Original Message-----
> > From: cisco-nsp-bounces at puck.nether.net
> > [mailto:cisco-nsp-bounces at puck.nether.net] On Behalf Of Sergio D.
> > Sent: Tuesday, November 14, 2006 9:14 AM
> > To: cisco-nsp at puck.nether.net
> > Subject: [c-nsp] MPLS MTU
> >
> > Hello folks,
> >
> > Some of our customers recently ran into some problems browsing and
> > pinging with over 1468. We fixed the browsing issue by adjusting MSS on
> > the incoming interfaces. It turns out that our core tag-switching
> > interfaces were set at the default 1500 which in turn allowed only a
> > 1496 or 1492 IP payload depending on how many labels.
> >  So now we are changing the MPLS mtu to 1508, but on Cisco's website it
> > mentions the following :
> > "Setting the MPLS MTU to a high number can lead to packets being dropped
> > on some devices, because the labeled packet is larger than the interface
> > physical MTU."
> >
> > A lot of our core is FE interfaces, I don't have a way to test this
> > before, wouldn't just automatically raise the physical MTU?
> > So after raising this an FE MTU goes to 1526? 1508 MPLS + 18 Vlan +
> > ethernet header?
> > Thanks,
> >
> > --
> > Sergio Danellli
> > JNCIE #170
> > _______________________________________________
> > cisco-nsp mailing list  cisco-nsp at puck.nether.net
> > https://puck.nether.net/mailman/listinfo/cisco-nsp
> > archive at http://puck.nether.net/pipermail/cisco-nsp/
> >
> >
> >
> 
> 
> -- 
> Sergio Danellli
> JNCIE #170
> _______________________________________________
> cisco-nsp mailing list  cisco-nsp at puck.nether.net
> https://puck.nether.net/mailman/listinfo/cisco-nsp
> archive at http://puck.nether.net/pipermail/cisco-nsp/

-- 
  ++ytti


More information about the cisco-nsp mailing list