Hmmm.... There's no reason you can't have flattening and
hiearchy at the same time, but let me state it another way
without hierarchy: You aren't going to fit the requirement that
routers need not be upgraded without fitting the requirement that
information hiding takes place.
Which moves the problem to information hiding, which is mostly
going to center around aggregation of information (unless you can
hide information while you still have it, but then that doesn't
make sense, because just having it means you have to handle it,
which then requires an upgrade as more information becomes
available). Aggregation is not a concept unique to routing
protocols as a means of reducing the density of information being
processed.
Anyway whether or not aggregation can occur without hierarchy is
another question, and perhaps it should be stated that way in the
requirements doc. First, I don't know that it can. A full mesh
between hierarchical units does not imply no hierarchy, and it
doesn't imply that aggregation cannot occur.
In fact, the entire concept of being able to have more than two
levels--mentioned elsewhere in the doc--implies levels, which
then implies hierarchy.
So, on to two sweeping statements:
-- If hierarchy is not to be presupposed, then eny references to
levels or pieces of the network being fit together to make larger
pieces, both indicative of hierarchy, should be removed, and we
should allow people to postulate a truly flat network, with
strict host routing between each host throughout (the absolute
result of no hierarchy, which is actually quite easy to achieve,
just replace every instance of a router in the Internet with a
switch).
-- If hierarchy seems to fall in behind everything we do, even
the advertisement of wires rather than reachable hosts, for
instance, then the presupposition of hierarchy at least at some
level should be mentioned. Perhaps even having a section on this
would be useful (?).
Just asking for consistency, that's all.
:-)
Russ
On Thu, 7 Mar 2002, Dmitri Krioukov wrote:
> > > > 4.5 IP Prefix Aggregation
> > > >
> > > >This section seems to contradict the earlier section which states
> > > >that the system must be scalable. :-)
> > >
> > > We're just saying that if you can come up with a way to scale
> > > without aggregating prefixes, that is just fine...
> >
> > Ummm.... Yes, well, good luck. A distributed database that
> > handles billions of entries and converges in milliseconds to
> > support voice, and does not have a constrained interconnection
> > model.... Well, maybe with a four deminsional model, where we can
> > simply calculate the required routes and transport the results
> > back in time through the routing system....
> >
> > Further, this contradticts with the requirement that routers not
> > be infinitely upgraded, in some way, I think.
> >
> > :-)
> >
> > Russ
>
> This (central, in my opinion) problem has been discussed
> so many times by now. Yes, indeed, we don't know today
> how to marry scalability to non-hierarchical routing.
> Note, however, that flattening of the Internet topology
> is the major reason why the current hierarchical routing
> architecture does not scale, why we're doing this work
> today, and why both drafts agree that we cannot presuppose
> anything about hierarchical nature of the NGA (the Next
> Generation Architecture rather than the National Gallery of Art :)
> Admit that it would be strange to state that aggregation/
> hierarchy is a MUST in the NGA, while the major reason
> why we're working on the NGA is that Internet topology
> flattening (as a given) contradicts with aggregation/
> hierarchy. Since we don't know how to do that today,
> I do not see any way other than "significant research
> effort."
> --
> dima.
>
__________________________________
riw@cisco.com CCIE <>< Grace Alone
This archive was generated by hypermail 2b29 : Mon Aug 04 2003 - 04:10:04 EDT