[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[RRG] Re: FYI -- Informal LISP BOF scheduled for lunch time on Thursday



On Wed, Dec 05, 2007 at 07:32:33PM -0800, Robert Raszuk wrote:
> Hi David,
>
> So today at the Routing Area Open Meeting you have explicitly expressed the 
> concern that "routing is about to collapse".
>

	I'm not sure what I said exactly, but if I said that, I
	probably misspoke (one of the dangers of speaking on the
	fly, I suppose). But in any event, there is stress on the
	routing system. And BTW, I'm sure one can find ISPs that
	will tell you, "nah, its not a problem", because they have
	the latest and great router [or whatever], but there are
	plenty of folks feeling the pain. If you dispute that,
	search, say, nanog@nanog.org (or just the archives of the
	last meeting). Convolve with the pressure the RIR system 
	is feeling (for example; there many factors here) and one
	can see storm clouds on the (near) horizion. 

	So what we are facing is not as unidimensional as being
	able to build a bigger router. We have plenty of
	existence proofs that we can do that. 
 
> I have been following and discussing with various folks this new wave of 
> fixing the routing. In fact I have my own set of ideas which are in fact 
> very very close to what Lixia was already proposing. Till now even looking 
> ten years ahead no one can show the significant proof that "routing is 
> about to collapse".

	Well, "about to collapse" probably isn't the way to frame
	it up. Its a complex problem, involving physics
	(dynamical systems), technology (what we are ostensibly
	working on here), economics, public policy (and
	associated regulatory environment), market dynamics,
	etc. I've done a bit of work on what the convolution of
	these various factors might mean, but this is still very
	much more an art than a science. See e.g.,  
	http://www.1-4-5.net/~dmm/talks/NANOG41/perfect_storm.pdf
	for some thinking along these lines.
	
> Contrary there are number of vendors with deployed platforms which are 
> today carrying ten fold number of routes what today's internet carry for 
> some other applications.

	We all know that we vendors say that. We can always build
	something bigger; that is simply not the point. IMO we
	can dispense with this. Its kind of a red herring (I do
	seem to recall someone saying "with enough thrust..." and
	yeah, I know it was Milo). So again, that one can build a
	bigger router is not the point.

	And BTW, as you know, precise models of how the routing
	system works (as a dynamical system) are not available,
	and while although we've done significant reverse
	engineering, to some exent we're all guessing about the
	properties of the system. But 

	we can say a few simple things, such as (i). the DFZ is
	growing, (ii). as the DFZ grows, more of the dynamics of
	the network are exposed in the core (simple computability
	arguments here), and (iii). the widespread deployment of
	IPv6 is a wildcard (O(2^32) >> O(2^128)). 

> PS. Said all of the above I think there are many other benefits for 
> introducing the hierarchy in the inter domain routing .. hence I am very 
> much supporting this as individual. I am just trying to make sure we focus 
> on the correct problem not the imaginary/non-existent ones :).

	That seems only prudent, and I'm sure everyone feels the
	same way.

	Dave

Attachment: signature.asc
Description: Digital signature