Re: [Diffserv-interest] QoS in diffserv network

[Brian E Carpenter <brc@zurich.ibm.com>]

> Zoltan.Balint@alcatel.be wrote:
> 
> Brian, Feng,
> 
> >> Can we always assume that a higher service class provides
> >> the same or a better service than any lower service class
> >> in AF?
> >>
> >>
> > No. There is no "higher" or "lower" built into AF. There
> > is simply the option of having one or more mutually
> > independent AF service classes, with an arbitrary number
> > of four service classes being named AF1, AF2, AF3, AF4
> > if you choose to use the recommended code points.
> >
> > If you happen to give each of these classes equal weights
> > in a WFQ scheme they are all offering the same level of service.
> >
> That would be the case when the relative load in those classes was equal as well. But, since some
> classes can be more loaded than others, the service seen in different classes does not
> relate to the index of the AF class. 

Zoltan,

There is no such thing as an index; the digit in AF3 is just part of the name.

If two AF classes have equal WFQ weights but different offered loads, they will 
experience different loss and jitter rates, but the service *offered* is the same.
The service *delivered* will vary, but that is a different matter. The two SLAs may 
be different, by allowing different levels of overbooking, but that is why the diffserv 
documents distinguish between the SLA and the SLS [service level specification].

I think we are in agreement, but I want to be precise.

> Unless, of course, the scheduler was strict priority, which does
> not conform to the requirements in RFC2597 (minimum bandwidth).

True.

   Brian

_______________________________________________
Diffserv-interest mailing list
Diffserv-interest@ietf.org
https://www1.ietf.org/mailman/listinfo/diffserv-interest