increasing the ethernet MTU greater than 1500 (1502)
kamal kc
kamal_ckk at yahoo.com
Fri Jan 6 18:35:01 PST 2006
--- Richard A Steenbergen <ras at e-gerbil.net> wrote:
> On Fri, Jan 06, 2006 at 10:02:07AM -0800, John-Mark
> Gurney wrote:
> > Nope, for pre-gige, only 1500 MTU is supported...
> This was extended
> > slightly to support vlan tagging, but I believe
> many of the drivers to
> > If you have a good gige card, you can go to 9000
> MTU and beyond...
> Technically speaking, the problem was not fixed by
> GigE or even 10GigE.
> The wonderful folks over at the IEEE feel there is
> no need to bring us out
> of the ethernet networking dark ages by defining any
> kind of standards for
> anything > 1500 bytes. We're left with a bunch of
> vendors who are each
> trying to do the right thing by picking random sizes
> ranging from 1518
> (stock + 4 bytes for a single .1q tag) to 16384, but
> there are no real
> standards, no mechanisms for ensuring
> interoperability, no protocols for
> negotiating MTUs between networks, etc.
Lots of newer FastE cards
> support jumbos or some kind
> of mini jumbo too. There are still plenty of NICs
> and switches out there
> with no or very half-ass jumbo support though.
thanks for providing the insight. I would try
with GigE next time.
since it seems a bad option to try MTU larger than
1500 in 10/100 Mbps ethernet i tried another option to
solve my problem.
Now i don't add additional header(2 bytes) to
distinguish
the packets that were processed. I rather tried for
protocol mapping.
This is what i did,
i mapped the protocol (in the protocol field of the ip
header)
0..56 to 138..194 if compressed
0..56 to 195..251 if uncompressed
with this simple protocol mapping i could only
compress
57 protocol data.
I guess the protocols 138 to 251 will not be used for
couple of years.
OR Is there any slighest possibility ????
thanks,
kamal
__________________________________________
Yahoo! DSL Something to write home about.
Just $16.99/mo. or less.
dsl.yahoo.com
More information about the freebsd-net
mailing list