QoS or not QoS; RE: B/W vs QoS Re: Any SPs using QoS ???

Nice lab theory. Where’s the practical application of the theory? The presentation did not discuss real-world implications and applications. The theories, at least as far as they can be discerned, do not take human error, differing hardware configurations, “weakest link” issues and a non-uniform universe into consideration. It also seems to indicate a constant bit-service demand.

One of the hardest parts about learning economics from a textbook is that it’s tough to know when to apply theory and when to use real-world examples. The presentation does not refute the original writer’s perceptions, nor, in fact, real world experience.

Mark Jasen

-----Original Message-----
From: owner-mpls@UU.NET [mailto:owner-mpls@UU.NET]On Behalf Of Ronaldo Moreira Salles
Sent: Thursday, September 28, 2000 3:27 AM
To: Jay Wang
Cc: mpls@UU.NET
Subject: Re: QoS or not QoS; RE: B/W vs QoS Re: Any SPs using QoS ???

I agree, but how can you support your arguments given the following technical analysis:

http://www.statslab.cam.ac.uk/~frank/TALKS/RoySoc99/sld009.htm

Cheers,

Ronaldo.

----- Original Message -----

From: Jay Wang

To: 'Grenville Armitage' ; mpls@UU.NET

Sent: Wednesday, September 27, 2000 8:17 PM

Subject: QoS or not QoS; RE: B/W vs QoS Re: Any SPs using QoS ???

I think Grenville has a point. To add to that, I think QoS
can provide at least the following that simply throwing lots
of BW at the network can not achieve:

* Service Performance Predictability (throughput, jitter, latency, etc)
* Service Prioritization
* Bandwidth Slicing
* Economical Scalability/Expandability
* Support of Targeted Service Demand Profiling and Projection

These QoS added values would have strong business implication to most ISPs.

- Jay

> -----Original Message-----
> From: Grenville Armitage [mailto:gja@research.bell-labs.com]
> Sent: Wednesday, September 27, 2000 11:32 AM
> To: mpls@UU.NET
> Subject: B/W vs QoS Re: Any SPs using QoS ???
>
>
>
>
> B/W isn't QoS.
>
> B/W may become cheap(er), but the premium service will
> be constrained jitter/loss (or predictable delay/loss, take
> your choice of poison).
>
> It isn't even clear that one can solely say "overprovision"
> as the magic words to clear up jitter and loss. Realities
> such as slow provisioning schedules and technology limitations
> can cause your B/W growth to fall behind the demand growth,
> creating congestion points.
>
> Which suggests anything that helps distribute congestion points
> (e.g. MPLS TE) and helps differentiate traffic at congestion
> points (e.g. Diffserv+MPLS) will find use.
>
> cheers,
> gja
>

The MPLS WG Archive

QoS or not QoS; RE: B/W vs QoS Re: Any SPs using QoS ???