Cell Relay Archive

[Mike Uttormark <muttormark@sntc.com>]

> Mike Uttormark wrote:
> 
> > The problem with using one lecs over multiple peer groups (if your
> > domain spans
> > multiple peer groups) is that it makes for one point of failure for
> all
> > the elans in
> > your entire domain.
> 
> Mike,
> 
> I think you're making the assumption that no ELAN is bigger than
> a peer group.  That may often be true but it certainly is possible
> to have them bigger than that (especially if you have special
> reasons for having particularly small peer groups).  If you have
> a multi-PG ELAN, you would also want advertisement of the LECS for
> that ELAN over the entire set of PGs it occupies.
> 
> 	paul
> 
I guess its hasn't been clear enough, so i will restate my assumptions:

A network with:
	a) 1 or more elans
	b) more than one peer group
	c) clients in multiple peer groups
	d) a single LECS at the well known address.

I say this is bad because it requires the single LECS to be visible in
multiple
peer groups, and is thus a single point of failure for all the elans in
all the peer
groups.  its much better to have an LECS in each peer group, configured
identically,
and restrict their visibilities to only one peer group each.

The PNNI MIB includes an object called pnniNodeDomainName:

	"The name of the PNNI routing domain in which this node
participates.  All lowest-level
	PNNI nodes with the same pnniNodeDomainName are presumed to be
connected."

As far as I know, this value of this object is purely for human
consumption, and plays no role
in routing decisions.  The definition does not preclude a domain
spanning multiple peer groups,
although I think this may have been the attempt.  In any event, I had
this object in mind when
I referred to a domain (above), and I further assumed that a domain
could have multiple peer groups,
although in reviewing the definition it seems that this may not be
valid.

Mike Uttormark
SNT, Inc. - www.sntc.com
I just work for them, I don't speak for them.