[96000] in North American Network Operators' Group
RE: Thoughts on increasing MTUs on the internet
daemon@ATHENA.MIT.EDU (Neil J. McRae)
Fri Apr 13 03:37:24 2007
From: "Neil J. McRae" <neil@domino.org>
Date: Fri, 13 Apr 2007 08:35:03 +0100
To: NANOG list <nanog@merit.edu>
Errors-To: owner-nanog@merit.edu
Saku Ytti wrote:
> IXP peeps, why are you not offering high MTU VLAN option?
> From my point of view, this is biggest reason why we today
> generally don't have higher end-to-end MTU.
> I know that some IXPs do, eg. NetNOD but generally it's
> not offered even though many users would opt to use it.
Larger MTU size was something I did some work on back in the FDDI days and =
the benefits are significant. More than just CPU improvements. Throughput a=
nd server performance increased substantially also. But FDDI and the like =
didn't come cheap so little interest at the time. At the LINX a few provide=
rs did run larger MTUs during those FDDI days. We did some testing with SRP=
/DPT in Stockholm and London also and again it worked well but again not ch=
eap. (we were looking at this for storage and exchange of cached content at=
the time)
Unfortunately I think the time where IXPs could make a difference might be =
past - and to make this happen tere needs to be more of a demand from the =
members of tose exchanges, its not just a case of turning on a vlan either =
the impact to the main fabric needs to be understood. Also atleast here in =
Europe many of the circuits into exchanges are Ethernet based also and I su=
spect many circuits into exchanges would require a lot of work to support J=
umbos. And then again lots of circuits into customer premise are Ethernet =
based now also some on GFP based SDH systems, some ATM and other whacko tec=
hnologies with dubious support for jumbo or larger frames.
Then there is the actual interface card support of large amounts of jumbos =
which in my experience is questionable based on a limitedl amount of testi=
ng though - Come back POS all is forgiven!
Regards,
Neil
