The MPLS WG Archive

["Dave Humphrey" <dave.humphrey@telindusk.net>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Juniper computes the ERO from the Traffic engineering database
the path signalled follows exactly that created by the ERO. If
it can't create it the LSP won't get set-up. JUNOS thus does
use the ERO.

Dave Humphrey

- -----Original Message-----
From: Giles Heron [mailto:giles@packetexchange.net]
Sent: 07 September 2001 13:46
To: Mareline Sheldon
Cc: mpls@UU.NET
Subject: Re: several questions


Mareline Sheldon wrote:
> 
> Please find my comments inline ..
> 
> ----- Original Message -----
> From: "Giles Heron" <giles@packetexchange.net>
> 
> > > topology into areas and so on.
> >
> > But the opaque LSAs used for TE don't cross area boundaries.  So
whilst
> > you can divide the network topology into areas you can't actually
> > do 
TE
> > across those areas.  But there is ongoing work on multi-area TE
> > to 
fix
> > this...
> 
> Don't we use opaque LSAs of link-state-type 11 in TE which are
> flooded throughout the AS ?

TE is carried in type 10 LSAs.  So only flooded through the area.

> > > Explained above. You can use route summaries .. and hence
> > > proves 
that
> MPLS
> > > is scalable ;-)
> >
> > Well, sort of.  Like I say, CSPF won't work across area
> > boundaries. 
> >
> > And if you are using LDP (not CR-LDP, that doesn't seem to exist
> > in 
any
> > approximation of reality) to signal your LSPs then route
> > summaries 
are a
> > bad idea, since you need the FECs to match prefixes in your IGP.
> 
> I am not really sure if CR-LDP doesn't seem to exist any more ..
Nokia,
> Nortel Networks and a lot of other giants are still actively
> involved 
in the
> development of CR-LDP. But then i may be wrong !

When was the last time you saw a router built by Nokia or Nortel?

They are hardly "giants" in the TCP/IP world.

For better or worse CR-LDP is dead (at least wrt MPLS.)  Can't speak
for
GMPLS.

> > > Please note that EXPLICIT_ROUTE object (ERO) is an optional
> > > objetc 
> contained
> > > in an RSVP PATH message. It is *only* when the ERO is present,
> > > the 
RSVP
> PATH
> > > message is forwarded towards the egress LSR along the path
specified by
> the
> > > ERO, *independent* of the IGP shortest path.
> > >
> > > Thus by default the path taken is the one guided by the IGP
running !
> >
> > But YMMV depending on the RSVP-TE implementation in your router.
Some
> > boxes always send an ERO I think?
> 
> I am not so sure if the boxes always send an ERO .. and i dont
> think 
Juniper
> does that .. atleast they dont claim to do so !
> btw whats YMMV ??

Your Mileage May Vary.

- From what I remember a certain other big router vendor always sends
an
ERO.  But since I don't have one in front of me to test on I can't be
definitive right now :)

> > > traffic is transferred to it before the old LSP tunnel is torn
down.
> >
> > I'm not sure that you have answered the question here?
> 
> Hmmm .. if there are no resources available at a *particular* LSR
> then 
why
> can rerouting be done ??
> I'm confused ..

But his question wasn't about re-routing?

It was about an LSP setup which hits a problem at an LSR which has
insufficient resources to complete the request.   In that case the
LSP
setup fails, and the head-end LSR has to re-signal.

Giles

- -- 
=================================================================
Giles Heron    Principal Network Architect    PacketExchange Ltd.
ph: +44 7880 506185              "if you build it they will yawn"
=================================================================

-----BEGIN PGP SIGNATURE-----
Version: PGP 7.0.4

iQA/AwUBO5jQ3NDzf3p4iWAnEQIYaACfXATZC0bNAJxtxPwp+wuMBEiGIdMAn3LX
i2R663upubi/2odETJMGyWmV
=bAuS
-----END PGP SIGNATURE-----