[c-nsp] setting max mtu on switch (Jumbo)

[Lee]

On 5/29/12, Gert Doering <gert at greenie.muc.de> wrote:
> Hi,
>
> On Tue, May 29, 2012 at 06:40:38PM +1000, Reuben Farrelly wrote:
>> I've often wondered about supposed downsides myself.  Why is it that we
>> don't see the layer 2 MTU set as high as possible on Cisco devices out
>> of the box, but a relatively "normal" routing MTU set to 1500 in the
>> default config?   Are there any "bad things" that could come out of this
>> config?
>
> I've posed that question here on the list about half a year ago, and
> nobody had to report anything negative...  so yes, I'm wondering as
> well why the default isn't "9200 off the shelf".

I was under the impression that all hosts on the vlan had to have the
same MTU or they couldn't talk to each other.  If that is the case,
having a switch default to 9200 byte MTUs opens up the possibility of
some users deciding they want large MTUs, giving it a try and, when it
works, setting their machines to 9200 byte MTUs.  Meanwhile, everyone
else of the vlan is still using a 1500 byte MTU..

Another issue in the "users doing their own thing" category is
following random "best practices" docs they find.   I was helping one
user who had pretty much all ICMP blocked in iptables because it was a
'best practice'

> One possible thought I had was that it could affect the internal memory
> layout, with a larger MTU segmenting the memory into fewer + larger
> buffers, but nobody was able to confirm that - at least not for the
> 6500...

Sounds reasonable.  I couldn't find anything for 6500s, but a quick
look found this for ASAs:
" You can enable support for jumbo frames for all interfaces by
increasing the amount of memory to process Ethernet frames. Assigning
more memory for jumbo frames might limit the maximum use of other
features, such as access lists."
  (http://www.cisco.com/en/US/docs/security/asa/asa82/configuration/guide/intrface.html)

Lee