The MPLS WG Archive

[Ping Pan <pingpan@cs.columbia.edu>]

Curtis Villamizar wrote:
> 
> In message <5.0.0.25.2.20000928083919.02f81ae0@flipper>, Fred Baker writes:
> > At 03:00 PM 9/26/00 -0400, Martin Picard wrote:
> > >   Service Providers tend to say that their backbone will never be
> > >   congested and if it ever gets there, then, bandwidth will be increased
> > >   and therefore no Congestion Management or Congestion Avoidance
> > >   mechanisms are necessary.
> 
> Thats clearly what the ISP marketing people have to say.
> 

Yes. 

>From various studies (Vern Paxson...), it was shown that end-to-end
jitter can be quite high. Within core networks, packet delay and loss
may be low. But after going through multiple domains, NAP's and POP's,
not sure end-user service quality can be assured. (Some time back, I saw
data shown that Internet packets go through 5-6 domains on average....
That's quite a few networks to travel.)

> The network also needs to degrade gracefully when massive outage
> occurs.  A few examples in the past 5 or so years (from memory)
> include earthquake in southern CA, CO fire in the Bay area, gas leak
> (power everything down) in LA area, floods in the mid-west, Amtrak
> wreck affecting both sides of track on East, gas company tearing up
> wrong pipe (major fiber conduit), and a hurricanes in the East
> affecting riverbed fiber, numerous smaller hurricane outages.  I'm not
> in operations so this is just a very small subset covering only very
> big outages.  The network can't just work well on sunny days.
> Congestion avoidance is needed.
>

Yep!

Another point here is the so-called Disruptive Innovation Effect (DIE).
Andrew Odlyzko of ATT Research had talked about that the Internet
traffic was doubled every 3-4 months in 1995 and 1996 due to the WEB
expansion. A similar traffic growth pattern was shown in some networks
last year sue to Napster. So unless SP's have some way (call it QoS
guarantees or Traffic Engineering or Flow Isolations...) to protect
existing users from sudden traffic bursts, I doubt that simply adding
bandwidth to the network is the solution. 

Finally, from all the conversations that I had with SP people, they all
complained about the complexity of managing QoS with the existing
protocols, tools and fancy architectures. I never hear any of them
saying that they don't need QoS.

> Curtis

- Ping