North American Network Operators Group

Date Prev | Date Next | Date Index | Thread Index | Author Index | Historical

Re: Points of Failure (was Re: National infrastructure asset)

  • From: Chris Woodfield
  • Date: Mon Sep 24 15:26:40 2001

One thing to keep in mind that the number of routes that disappeared from
the routing table, while a start, is in no way the final arbiter of how a
particular incident affected the internet's performance. While the direct
effect is obvious, other, less easily measurable consequences can result
from the loss of capacity producing congestion on the remaining network
infrastructure, as well as increased latency resulting from longer
backhauls. If there's plenty of available backup bandwidth, this effect
wil be minimal, but if not, let the pain begin. 

As an example, a certain DSL provider has had to reroute their Covad PVCs
from 25 Broadway  to their Boston and Washington, DC installations, resulting 
in their capacity to Covad in those locations being almost completely
pegged 24/7. Not fun to be their customer in any of these areas right
about now.


On Mon, Sep 24, 2001 at 12:22:08PM -0700, Grant A. Kirkwood wrote:
> Sean Donelan wrote:
> > 
> > On Mon, 24 Sep 2001, Bob Bownes wrote:
> > > But there was a point in time when taking out a certain parking garage
> > > in Va could have caused us a very great deal of difficulty. But I'd say
> > > we are past that, for the most part.
> > 
> > Are we?
> > 
> > When 25 Broadway failed, approximately 1% of the global Internet
> > routing  table also disappeared.  Which I would guess qualifies it
> > as a "major" hub.
> But does that mean that X number of sites were unreachable, or that
> there were simply Y number fewer routes to X sites? (Excluding those
> *directly* affected, ie; those *in* 25 Broadway)
> > Verizon still has 100,000 lines out of service, and only now
> > begun to restore service to "small" businesses.
> Yes, but my understanding was that we were referring to IP traffic. POTS
> doesn't exactly have a built-in routing protocol.
> > A couple of years ago a fiber cut in Ohio disrupted about 20% of
> > the Internet routing table.
> But again, does this mean that 20% of the Internet was unreachable, or
> that there were 20% fewer routes to a given number of (hopefully
> multihomed) sites?
> No, this question is not rhetorical... I simply don't have any imperical
> evidence to look at that could adequately answer this question.
> Grant
> -- 
> Grant A. Kirkwood - [email protected]
> Chief Technology Officer - Virtical Solutions, Inc.