[c-nsp] What MTU for Bellsouth BBG / BRAS <-> LNS l2TP tunnel?

Ash Garg ash at telstra.net
Wed Oct 27 21:42:14 EDT 2004


I decided on 1560 as follows:

Assuming 48 bytes of payload for ATM Cell and total of 32 transmitted cells
means total octets sent = 1538 and not enough to prevent l2tp fragmentation.
If we pick 33 cell then total octets sent is 1584. So an IP mtu of 1560 + 8
bytes of AAL5 Header + 8 bytes of AAL5 Padding + 8 bytes of AAL5 Trailer =
1584.

This gives me 10-12 bytes "extra" in the IP packet and 8 bytes "extra" in
the AAL5 frame which covers any weird stuff that may occur and does not "I
think" put extra load on the LNS.


Regards,
Ash


-----Original Message-----
From: Brian Feeny [mailto:signal at shreve.net]
Sent: Thursday, 28 October 2004 11:11 AM
To: ash at telstra.net
Cc: cisco-nsp; Jon Lewis
Subject: Re: [c-nsp] What MTU for Bellsouth BBG / BRAS <-> LNS l2TP
tunnel?




I see so your trying to allow for the larger payload?  I would think
1548 would nail it assuming
worst case l2tp/pppoE, etc.  Maybe 1550 if your using sequencing
options.

What really interesting is to look at the entire encapsulation process
and calculate the most
"ideal" MTU that would lead to the most efficiency (less fragmenting,
and more efficient use of
packets/cells).

Brian


On Oct 27, 2004, at 7:49 PM, Ash Garg wrote:

> Guys, we have set to 1560 on the ATM subIF and haven't had any
> problems with
> MTU so far.
>
> Regards,
> Ash
>
>
> -----Original Message-----
> From: cisco-nsp-bounces at puck.nether.net
> [mailto:cisco-nsp-bounces at puck.nether.net]On Behalf Of Jon Lewis
> Sent: Thursday, 28 October 2004 10:40 AM
> To: Brian Feeny
> Cc: cisco-nsp
> Subject: Re: [c-nsp] What MTU for Bellsouth BBG / BRAS <-> LNS l2TP
> tunnel?
>
>
> On Wed, 27 Oct 2004, Brian Feeny wrote:
>
>> So you are running MTU 4470 on the PVC over which your L2TP session is
>> built, but I am using 1500, and thats probably the difference.
>
> Yes.  The ATM interface and L2TP carrying PVCs have the default (for
> ATM
> DS3) MTU of 4470.  I haven't tried setting it to anything else to see
> if
> other sized would work / break things.
>
>> In the old days I use to think PPPoE was more ideal, because it could
>> terminate on the customer desktop (WinPoet, etc).  But of course DSL
>> is
>> almost always terminated into CPE, so maybe I should look more closely
>> at PPPoA vs. PPPoE.
>
> PPPoE may have been nice back when DSL bridges were so popular, but
> since
> pretty much all DSL CPE can do PPPoA now, may as well use that when
> possible.
>
> ----------------------------------------------------------------------
>  Jon Lewis                   |  I route
>  Senior Network Engineer     |  therefore you are
>  Atlantic Net                |
> _________ http://www.lewis.org/~jlewis/pgp for PGP public key_________
> _______________________________________________
> 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/
>
>
---------------------------------------------
Brian Feeny, CCIE #8036, CISSP
Network Engineer
ShreveNet Inc.




More information about the cisco-nsp mailing list