On Jan 15, 2016, at 9:49 AM, Robert Armstrong <bob
at jfcl.com> wrote:
Let me give you the same example I just gave
Peter in a private mail.
The problem with your example is that you have one area (1 in your case) that has more
than one area router AND all of them have external links. Worse, some of those multiple
area routers for area 1 have links to the same external machine. The problem is just a
badly thought out network topology. If area 1 had only one gateway to the external world,
then all would be well.
Now you're going to tell me "yes, but I don't want to do it that way".
That's fine, but like I said - the problem isn't with the technology.
But that is NOT an example of a bad topology. It is what a highly fault tolerant topology
would look like. You can certainly build star topologies, as you suggest, but you can
also design topologies that tolerate any single failure. Or even any double failure.
Single fault tolerant topologies would have two area exit routers, and redundant paths
within the area. That second aspect is critical because of the DECnet Phase IV rule that
in-area traffic never takes any out-of-area paths, even if the area is partitioned.
In a fault tolerant network, even if the circuit costs are symmetrical, the paths may not
be. In DECnet, because of the tie breaking rule, they will at least be predictable. IP
does things differently: it might use multiple paths if they have equal costs, and it may
choose among them in ways that are not specified (I'm not sure of the details of the
latter case). In DECnet Phase V, there does exist equal cost path splitting. I think it
was also added in Phase IV-plus, but I don't have any documents for that version. It
isn't in plain Phase IV, though the NSP spec does describe hanging on to out of order
packets (which is what you want to do if your network performs equal cost path
splitting).
paul