[e2e] "PMTUD using options" draft

[Michael Welzl]

Dear Ran,

I don't know much about routers - I can only replicate
what I've been told. I apologize for my claim, as I
remember that you said this before - if you look
at the web page, you will see that this:

Packets with IP options are known to be processed in the so-called "slow
path",

has just changed into this:

Packets with IP options are said to be processed in the so-called "slow
path" by some routers,

:)


The point of the web page + measurements is:

Some things may really make sense as an IP option, but
you usually can't propose them in the IETF because a
number of people tell you that this is unrealistic
because of slow path processing, which makes everything
at least 700 times slower.

No proof, no numbers, no measurements.

The idea was just to change that. If I know that IP
options typically lead to an approximate additional
delay of about 7%, I may be able to judge wheter a
proposal based on IP options makes sense or not.

I mean, I can't measure everything ... for instance,
what if I flood a router with packets carrying options?
Will the result be the same?

Still, I think that the measurements can at least
give you a rough idea of the order of magnitude,
which is better than nothing.

Best regards,
Michael






On Thu, 2004-02-12 at 15:54, RJ Atkinson wrote:
> On Feb 12, 2004, at 08:55, Michael Welzl wrote:
> > A local copy + some measurement results on the overhead
> > caused by option processing in routers can be found at:
> > http://www.welzl.at/research/projects/ip-options/
> 
> It is widely believed that all routers process IP packets
> with options in the "slow path", as you claim in your web
> site above.  This should be re-filed in the basket of
> "widely believed fallacy". :-)
> 
> In fact, this is no longer true (if it ever was).  It is true
> for certain widely used routers that rely on CPU-based forwarding.
> CPU-based forwarding is probably dying out in routers not sold
> in retail consumer electronics stores (and maybe even there).
> 
> However, many other firms have hardware-based forwarding engines
> that include the ability to process packets containing IP options
> in the hardware fast path.  Some of the better hardware-based
> forwarding engines are programmable, to varying degrees, so
> adding support for new IP options in the hardware fast path
> might be as simple as putting new microcode into a router
> system software image.
> 
> Whether such implementers would be motivated to implement
> support for any particular new IP option is probably driven
> mainly by economics.  Most firms have long lists of new features
> that customers desire and finite development resources.
> 
> Yours,
> 
> Ran Atkinson
> rja at extremenetworks.com
> 
> Disclaimer: Employed by, but not speaking for, Extreme Networks.
-- 
Michael Welzl <michael.welzl at uibk.ac.at>
University of Innsbruck