[nsp] Intelligent Routing Devices ?

[Stephen J. Wilcox]

On Tue, 20 Jan 2004, Jim Devane wrote:

> I am wondering about the future of "intelligent routing devices" and how
> relevant the greater Internet community believes them to be. 

Hmm these vendors have been quiet since the 'dot com crash'! Perhaps thats an 
indication of their relevance in that the market wiped out bad businesses, or 
maybe they just stopped calling me :)

> I am not interested so much in who's got the better methodology etc. so much
> as I am in their very relevance. Aside from marketing, is their presence
> likely to solve anything that needs solving?

You didnt ask if they might break something.. adding complexity to something 
that isnt broken is surely asking for trouble. I dont think finance is a 
constraint any more which was a big selling point to allow you to better utilise 
links (circuits + transit costs are small now)

> That is to ask, is the latency currently in the Internet originated from
> disaster ( back-hoes, black-outs etc. ) or is it created at peering
> points/NAPs/Exchange points ?

What latency? 99% of the time you can classify a disaster as having 'no effect'
or 'broken', my experience is that 'degraded' rarely occurs in backbone systems.

I'm not aware there are any issues with peering.. are you?

> Certainly there are both, and of course the former will exacerbate the
> latter, but on average, are NAPs still a significant problem that needs to
> be solved by intelligent routing? Or has the presence of Exchange Points
> alleviated enough of that congestion? 

I would be worried that the best solution to a 'disaster' outage is not to bring 
in levels of complexity, you're surely building a house of cards with your 
network already being in an unpredictable state and a whole bunch of unknowns 
getting fed into your IRD. The simple solution is your existing routing 
protocols and sufficient redundancy to take outages up to your desired level of 
broken-ness!

> I just cannot see and IRD solving disaster related latency, either the fiber
> cut is too small or too big to have any effect. However, if the Internet is
> still largely congested at the Interconnection points then I can buy the
> case for needing to thread my way through it.

See above, I didnt know it was congested at all, in fact people have more 
bandwidth than they know what to do with.

I dont mean this in a bad way but what is your source for your assertions, they 
dont (to me) look to be correct..

> ANY and ALL comments welcome. 

Steve