"Mcdysan, David E" writes:
> I had a general statement about scaling in the wg slides, and I recall
> Alex objecting that in general the IETF avoids quantitative values.
> This is an important issue, and is one we need to address to better
> understand the cost/benefit of approaches that differ from just
> extending the set of advertised parameters and/or their useage in
> signaled LSPs.
> I propose that we discuss a parameteric model, e.g., Number of nodes
> (N), total number of Links (L), some range of Component links per
> Composite (C(i)), frequency of latency changes (L), frequency of
> capacity changes, etc. and try to compare the "scalability" and
> "response time" in a parametric way. Then, outside of the IETF,
> quantitative numbers could be inserted.
Really thorough treatment of the topic could go to SigComm or any of a
number of places. A quick analysis or a summary of deployment
experience can go to NANOG.
The model can be proposed outside IETF as well. It makes no sense to
put it in the requirements document. In most analysis of IGP
convergence or CSPF workload or convergence there have been
assumptions about average degree of a node, etc. Models will vary.
> > -----Original Message-----
> > From: [email protected] [mailto:[email protected]]
> > On Behalf Of Tony Li
> > Sent: Tuesday, March 30, 2010 3:18 AM
> > To: [email protected]; John E Drake
> > Cc: [email protected]
> > Subject: Re: composite link - candidate for respin, maybe
> > >> A clear statement as to what we are trying to accomplish wrt IGP
> > >> scaling by using bundled links would be good.
> > >
> > > Please indicate what in the requirements we are discussing is not
> > > clear rather than a blacket statement that the goals are not clear.
> > Once again: what are the scalability goals that we're trying
> > to achieve, both on a per-system and network-wide basis?
> > This should be a quantitative requirement.
> > Tony
> > _______________________________________________
> > rtgwg mailing list
> > [email protected]
> > https://www.ietf.org/mailman/listinfo/rtgwg
rtgwg mailing list