Re: Just shy of OT MPLS Question

Same thing with Cisco as with Juniper. TE is based on the
'seeing-the-whole-network' idea so the algorithm (CB-SPF) can choose which
path it wants to take throughout the network. Therefore you are required to
have a link-state protocol as IGP for it to work. Another way would be doing
static path specifying each next-hop, then RSVP would go down that path to
make bandwidth reservation and allocate labels.

So no difference here between Cisco and Juniper when it comes to how TE works
:-)

--
Regards,
Rick Mur
CCIE2 #21946 (R&S / Service Provider)
Sr. Support Engineer  IPexpert, Inc.
URL: http://www.IPexpert.com
On 23 dec 2009, at 15:59, <Keegan.Holley_at_sungard.com>
<Keegan.Holley_at_sungard.com> wrote:
> Again just curious.  I know (at least with Junipers) the traffic engineering
database is based on the ospf/isis link-state DB's.  I was just curious if
there was some way to track TE metrics in BGP as well.  As far as the diffserv
tunnels that's one way of doing it.  With Juniper mpls it's more common to see
RSVP and bandwidth allocation and queueing done via resource reservations when
the LSP is formed.  Then you don't end up with a bunch of tunnel interfaces.
>
>
> Thanks again for your input,
>
>
> Keegan
>
>
>
>
> From:	Rick Mur <rmur_at_ipexpert.com>
> To:	"<Keegan.Holley_at_sungard.com>" <Keegan.Holley_at_sungard.com>
> Cc:	ALL From_NJ <all.from.nj_at_gmail.com>, Cisco certification
<ccielab_at_groupstudy.com>, nobody_at_groupstudy.com, Scott Morris
<smorris_at_ine.com>, JR Garcia <ttuner_at_gmail.com>
> Date:	12/23/2009 07:26 AM
> Subject: 	Re: Just shy of OT MPLS Question
> Sent by:	<nobody_at_groupstudy.com>
>
>
>
>
> I don't really see the reference to BGP, IGP, L2/L3VPN and TE :-) They all
are
> different and address a different feature. I know a lot of implementations
of
> TE. As of management purposes it's ridiculous to use TE for ALL your MPLS
> traffic (carrying your L2/L3VPNs), but for specific implementations it's
the
> ONLY way to go.
>
> The latest trend is DiffServ Tunnels. Sending traffic through a TE tunnel
> based on DSCP and therefore giving voice and video traffic preference
> throughout the network.
>
> (Some of the largest ISP's (and therefore the largest networks) in the
world
> use TE in their whole network, also the new Carrier Ethernet (or GMPLS)
thing
> that is going on, is quite based on the same principle)
>
>
> --
> Regards,
>
> Rick Mur
> CCIE2 #21946 (R&S / Service Provider)
> Sr. Support Engineer  IPexpert, Inc.
> URL: http://www.IPexpert.com
>
> On 23 dec 2009, at 12:58, <Keegan.Holley_at_sungard.com>
> <Keegan.Holley_at_sungard.com> wrote:
>
> >
> > Sorry just making sure.  So I suppose TE is out for our little scenario
> here.  L3/L2vpn?   Are there any real world uses for this or does it go in
the
> stupid router tricks file?
> >
> >
> > From:                 Rick Mur <rmur_at_ipexpert.com>
> > To:                 <Keegan.Holley_at_sungard.com>
> > Cc:                 ALL From_NJ <all.from.nj_at_gmail.com>, Cisco
certification
> <ccielab_at_groupstudy.com>, nobody_at_groupstudy.com, Scott Morris
> <smorris_at_ine.com>, JR Garcia <ttuner_at_gmail.com>
> > Date:                 12/23/2009 06:52 AM
> > Subject:                  Re: Just shy of OT MPLS Question
> >
> >
> >
> >
> > I think I answered the question in my previous mail :-)
> >
> > You CAN use BGP to get rid of an IGP, but you CAN'T use LDP then, LDP
uses
> an IGP protocol to allocate labels. If you want to allocate labels for BGP
> prefixes, you should use BGP as label distribution as well (neighbor
> send-label command).
> >
> > On the other hand, I would never implement this. An IGP with LDP is far
more
> easier to setup and maintain than using BGP for this purpose.
> >
> >
> > --
> > Regards,
> >
> > Rick Mur
> > CCIE2 #21946 (R&S / Service Provider)
> > Sr. Support Engineer  IPexpert, Inc.
> > URL: http://www.IPexpert.com
> >
> > On 23 dec 2009, at 12:43, <Keegan.Holley_at_sungard.com> wrote:
> > That's interesting.  I think that was my original question whether you
could
> build your base routing table using bgp and have LDP/RSVP base it's label
> assignments on it.  I knew you could do it with statics (routes not labels
> sorry..)  So just so I'm sure I understand you're saying that there's no
way
> to replace your IGP with bgp even if the next-hop and other issues are
taken
> care of?  I was just curious since I've seen some of the smaller carriers
> where every router is a PE router so you cannot get away from doing BGP in
the
> core.  I was just curious if there was a way to keep from doing the IGP in
the
> core and maybe save some resources that way even at the expense of TE.
> >
> > From:                 Rick Mur <rmur_at_ipexpert.com>
> > To:                 <Keegan.Holley_at_sungard.com>
<Keegan.Holley_at_sungard.com>
> > Cc:                 ALL From_NJ <all.from.nj_at_gmail.com>, Cisco
certification
> <ccielab_at_groupstudy.com>, nobody_at_groupstudy.com, Scott Morris
> <smorris_at_ine.com>, JR Garcia <ttuner_at_gmail.com>
> > Date:                 12/23/2009 03:00 AM
> > Subject:                   Re: Just shy of OT MPLS Question
> >
> >
> >
> >
> >
> > The IGP label is usually the 'outer' label (unless another label get's
> attached in a CsC environment). Inner label is the VPN label, Outer label
is
> the IGP label.
> >
> > A flat MPLS network with statics? I would never ever build a network with
> static label assignments :-) (doesn't even work on Cisco, it does on
Juniper)
> >
> > You can't use BGP and LDP to assign labels. If you want BGP prefixes to
get
> a label allocated, you do this within BGP and then BGP advertises and
assigns
> labels for those prefixes. This will definitely work. So NO LDP then :-)
> >
> > --
> > Regards,
> >
> > Rick Mur
> > CCIE2 #21946 (R&S / Service Provider)
> > Sr. Support Engineer  IPexpert, Inc.
> > URL: http://www.IPexpert.com
> >
> > On 23 dec 2009, at 03:52, <Keegan.Holley_at_sungard.com>
> <Keegan.Holley_at_sungard.com> wrote:
> >
> >
> > <nobody_at_groupstudy.com> wrote on 12/22/2009 02:36:40 AM:
> >
> > > I'm also a bit confused with the term MPLS Tunnels. Do you mean a
> > > normal MPLS VPN or a TE Tunnel, since both require totally different
> things.
> > >
> > Neither, I'm not that familiar with the term.  I always took tunnel to
mean
> anything encapsulated in something else and thus interchangeable with LSP.
In
> other words I meant LSP's.
> >
> > > For IBGP it's not really recommended and same as Bryan said, I never
> > > tested this. It does is an implementation for Inter-AS
> > > configurations to use the BGP send-label command to advertise
> > > prefixes and labels in BGP, which works :-)
> > >
> >
> > Are these inner or outer labels? Inner with one outer to keep you from
> having to run LDP with a foreign router? I would assume they are the inner
> labels and then your RSVP/LDP protocol would take care of the outer labels
> used to reach the next AS's router.
> >
> > > For TE tunnels, you cannot allocate labels through BGP, but solely
> > > through RSVP and a link-state protocol. This is because of he
> > > dynamic behavior and the SPF protocol that TE uses to calculate
> > > paths throughout the network. A distance vector protocol like BGP
> > > would not work as within a link-state you have the full topology of
> > > the network available to calculate your path on (within an area of
> course).
> >
> > Throw TE (and usefullness with it...) out the window for a minute.  For
> example you can create a flat MPLS network using static routes.
> >
> > >
> > > So for MPLS VPN's this should work, advertise the PE loopbacks in
> > > IPv4 BGP and advertise VPN labels in VPNv4 BGP. You should run BGP
> > > on the P of course then. Cool thing to lab it up :-)
> >
> >
> > Could you use BGP as the only protocol and just enable LDP on each
interface
> to do the switching?  Assuming next hop reachability was taken care of
> (without static routes).  Don't have any SP stuff setup right now or I'd
try
> it out myself.
> >
> > >
> > >
> > > --
> > > Regards,
> > >
> > > Rick Mur
> > > CCIE2 #21946 (R&S / Service Provider)
> > > Sr. Support Engineer  IPexpert, Inc.
> > > URL: http://www.IPexpert.com
> > >
> > > On 22 dec 2009, at 06:29, Scott Morris wrote:
> > >
> > > > Yup, you can.
> > > >
> > > >
> > > >
> > > >
> > > > *Scott Morris*, CCIE/x4/ (R&S/ISP-Dial/Security/Service Provider)
> #4713,
> > > >
> > > > JNCIE-M #153, JNCIS-ER, CISSP, et al.
> > > >
> > > > JNCI-M, JNCI-ER
> > > >
> > > > evil_at_ine.com
> > > >
> > > >
> > > > Internetwork Expert, Inc.
> > > >
> > > > http://www.InternetworkExpert.com
> > > >
> > > > Toll Free: 877-224-8987
> > > >
> > > > Outside US: 775-826-4344
> > > >
> > > >
> > > > Knowledge is power.
> > > >
> > > > Power corrupts.
> > > >
> > > > Study hard and be Eeeeviiiil......
> > > >
> > > >
> > > >
> > > > JR Garcia wrote:
> > > >> pretty sure you can exchange labels via ibgp using the "send-label"
> > > >> command. Eg: neighbor 1.1.1.1 send-label
> > > >
> > > >
> > > > Blogs and organic groups at http://www.ccie.net
> > > >
> > > >