[96001] in North American Network Operators' Group
RE: Thoughts on increasing MTUs on the internet
daemon@ATHENA.MIT.EDU (michael.dillon@bt.com)
Fri Apr 13 05:06:29 2007
Date: Fri, 13 Apr 2007 10:03:53 +0100
In-Reply-To: <461EF241.2090700@satchell.net>
From: <michael.dillon@bt.com>
To: <nanog@merit.edu>
Errors-To: owner-nanog@merit.edu
> No, I doubt it will change. The CRC algorithm used in Ethernet is=20
> already strained by the 1500-byte-plus payload size. 802.3=20
> won't extend=20
> to any larger size without running a significant risk of the CRC=20
> algorithm failing.
I believe this has already been debunked.
=20
> From a practical side, the cost of developing, qualifying,=20
> and selling=20
> new chipsets to handle jumbo packets would jack up the cost of inside=20
> equipment. What is the payback? How much money do you save going to=20
> jumbo packets?
I believe that the change is intended to apply to routers and the
ethernet switches that interconnect them in PoPs and NAPs and exchange
points. Therefore the cost of a small chipset modification is likely to
be negligible in the grand scheme of things.
As for numbers, it is not dollar figures that I want to see. I would
like the people who have jumbo packets inside their end-user networks to
run some MTU discovery and publish a full MTU matrix on all paths on the
Internet. That way we can all see where there is end-to-end support for
large MTUs and people who want to make buying decisions on this basis
will have something other than vendor assurances to show that a network
supports jumbograms.=20
--Michael Dillon
