"Mcdysan, David E" writes:
> > -----Original Message-----
> > From: [email protected] [mailto:[email protected]]
> > On Behalf Of Tony Li
> > Sent: Tuesday, March 30, 2010 12:27 PM
> > To: John E Drake; [email protected]
> > Cc: [email protected]
> > Subject: Re: composite link - candidate for respin, maybe
> > Hi John,
> > > If we are placing flows on the component links of a
> > heterogeneous link
> > > bundle, we may have to look at additional fields within
> > packets and/or
> > > maintain additional state, relative to what is required to
> > place flows
> > > on the component links of a homogeneous link bundle. The
> > example that
> > > has been given is flows that require low latency.
> The current draft combines change frequency and low latency. I think we
> need to separate these out as separate characteristics.
I don't see where that comment is coming from. There is nothing to
say that change frequency and low latency are combined in one
Please indicate what text you would like to see changed.
> > If we go by the definition that we're working with, a "flow"
> > is atomic and therefore should never be demuxed, just to
> > ensure ordering.
> > > I was agreeing with Tony, so please direct this question to him.
> > > (But, I don't see any real discussion of IGP scaling requirements.)
> > I think that this has to come from Dave. If I get to set the
> > requirements,
> > I promise a Very Low Bar. ;-) ;-) ;-)
> I agree that we need to have scaling requirements. Also, I think that
> convergence time (or response time to a transient) is something we need
> to specify. See my earlier note.
We can't specify convergence time. For one thing, what are we
specifying convergence of? Is it convergence of a load balancing
operation? If so, how fast are the offerred loads of the flows
changing? Is it convergence of an entire network after a fault? If
so what is the topology of the network that is converging and what
fault are we covering? Do we try to add realistic control plane
processing times? If so who's implementation?
> > Tony
rtgwg mailing list