On 05/09/07, paul at clubi.ie <paul at clubi.ie> wrote:
> On Wed, 5 Sep 2007, Thomas Bridge wrote:
> > No I don't - you're either not understanding or I'm not making
> > clear - there typically isn't an issue buying the international
> > transit required to support a service.
> Ok, then I don't understand. What was the issue with external links,
> can you expand on that?
>>> b) The ISPs further do not differentiate on cost of bandwidth. So a
>>> set of customers who use near-100% of their link bandwidth to talk
>>> to each other, but only a trivial amount of an ISPs expensive
>>> transit bandwidth, are penalised the same as customers who use
>>> 100% of their bandwidth to share data with hosts /external/ to the
and suggested that it was a "massive problem".
In fact, it isn't. Even in the context of running P2P networks,
which is where we started, the biggest cost issue is typically getting
the traffic into the data centre itself.
> > I think you're getting so obsessed with network purity, you're
> > bemoaning the fact that MPLS, L2TP and BGP are all protocols
> > designed to deal with different administrative domains.
> No I'm not. I'm bemoaning the fact we invented MPLS and L2TP -
> because of shortcomings with IP, and (to a lesser extent) because of
> how telecommunication networks had to deal with deregulation, across
> the world.
L2TP possibly (in fact - AIUI, it started off as a protocol to allow
telcos to offer wholesale dialup options in the USA to mom and pop
ISPs who wanted to expand their footprint without investing in the
circuits and equipment needed to go to new areas).
> I'd disagree, the current model is one where telcos and others have
> done all they could to avoid facing the premise of IP (simple) and
> packet-switching generally (independent, but informed,
> hop-by-hop/packet-by-packet routing).
> That's an anathema to telcos. To avoid it, they've helped develop
> several circuit-switching transports, so they can continue to pretend
> to customers their networks consist of independent circuits..
I think you are confused here - most of the circuit switching
transports predate IP. Go back 15 years and you'll find plenty of
> > Economies of scale. And it's the nature of many industries that
> > they start off with lots of smaller companies and then finally
> > consoldidate into many big ones.
> Yes, economies of scale means consolidation. Note though that NOT A
> SINGLE reseller of wholesale RADSL has survived (ie Netsource).
Digiweb are still independent as I recall?
> Though, there weren't many independent ISPs in Ireland to begin with.
> There are a lot more in Britain, but there the market is biased by
> the general aversion to end-user customer service of the incumbent
That would be same incumbent who recently announced they had 4 million
> >> Anyway, no matter the business model: Scenic routing sucks..
> > It's the nature of the world we live in. Its the nature of networks
> > that traffic doesn't always take the physically shortest path.
> It's the nature of networks that they /should/ route traffic
> according the shortest path in the topology, according to some
> arbitrary metric.
That doesn't contradict what I said.
The issue is you don't like the arbitary metric being used.
> It's also obvious that todays consumer broadband networks fail (by
> quite a margin) to do so. Unless of course you think "political
> expediency" is something that ought to be factored into network
> design ;).
Yes there are inefficencies in the design of current consumer
broadband networks. But the inefficencies in terms of routing from
number 16 Acacia Avenue to number 17 only applies to a very, very,
very small percentage of traffic.
So it's a trade off between a pure network design and straight forward
management of the network. (Cable networks are slightly different if
they use ethernet in the street).
> The problem with not taking the shortest-paths, instead ignoring them
> in favour of routing via some small set of central points, is that
> you get large loads at those central points, loads which NEED NOT be
> there, if you used the network fully.
So what? That's generally regarded as good network design - bring
traffic towards the centre. Ideally traffic within somewhere like
an exchange would stay within the exchange, but the amount of
additional load caused by bringing it to the centre is insignificant.
> > At the risk of starting a whole new thread by it self - how is IPv6
> > "less costly" than IPv4 to administer?
> It's capable of auto-configuration.
> > But why? Other than aesthetics?
So instead of taking 1518 bytes to send your 1500 byte packet, you now
use 1568. That's an increase of 2.6% - I'll live with it.
> > My viewpoint is hardly unique - the entire point of an IP layer is
> > that I route IP - I don't worry about what the guys at the lower
> > layers do. In the broadband aggregation world for instance, I
> > simply want a lot of point to point links to my client sites.
> And why do you want that?
Ease of management. Duh!
> > In $LASTJOB, we had a clear and distinct management domains between
> > the various layers of the network - we had the fibre guys, the SDH
> > guys, and the IP guys. We didn't worry too much about what the
> > other layers were doing.
> Right, and you shouldn't.
> Except you didn't have that nice layering. You had:
> Fibre/SDH guys:
> - the layer1 guys (wire pullers ;) )
> - the layer2 guys
> - the layer3 guys (ATM I guess?)
No ATM there.
> FWIW, there are ways to do multi-topology, shortest-path routing
> correctly (ie efficiently).
Do explain how it works in a situation that allows the consumer the
choice of ISP and has clear management domains for the ISP?
> We ought to be doing IP routing between them at either the exchange
> (IP routing capable DSLAMs are available, even if most today are
> not), or else at the BAS (which are IP routers).
Which can only be done by having one ISP.
> > And when Magnet and BT unbundle the exchanges in Limerick, and stop
> > using the BAS, the traffic will still travel all the way around
> > Ireland so that they can cross over to Eircom at INEX.
> Yes, this is known as a bad thing.
Which can only be fixed by having one ISP.
> Then that's a problem with IP, which ought to be fixed. Or
> alternatively, we should have switched the internet to native ATM..
I think that ship has sailed.
> > Yes - it's designed to ensure that instead of taking the
> > technically best path, it takes the path most appropriate based on
> > administrative domains, cost and politics.
> Right, that's known as the 'best path'. The metric is set arbitrarily
As opposed to being the best path you were talking about in the
consumer DSL passage above.
> You can not apply routing policy to links which do not exist.
BGP of course isn't link aware either.
> > Of course, the fact remains that one of the reasons the ISPs aren't
> > particularly interested in fixing that problem is that the amount
> > of traffic that is exchanged between users within the same exchange
> > is tiny.
> Times a couple of hundreds and it all adds up.
0.1% is still 0.1% - regardless of whether it's of 64k or 10G.
> Problem today is, to do it today would be a nightmare cause:
You can't have more than one ISP if that's how you want to do it.
Maintained by the ILUG website team. The aim of Linux.ie is to
support and help commercial and private users of Linux in Ireland. You can
display ILUG news in your own webpages, read backend
information to find out how. Networking services kindly provided by HEAnet, server kindly donated by
Dell. Linux is a trademark of Linus Torvalds,
used with permission. No penguins were harmed in the production or maintenance
of this highly praised website. Looking for the
Indian Linux Users' Group? Try here. If you've read all this and aren't a lawyer: you should be!