[c-nsp] BGP Signalled VPLS

Wed Apr 24 11:00:32 EDT 2013

> Similarly if PE-C has active PW to both PE-A and PE-B then when it sends a
BUM frame across the VPLS, the CE-A will receive the BUM frame twice.
Ah yes right I agree. 
I drove myself a bigger picture and now it makes sense. So the redundant PE
is not connected to anyone it just sits there waiting to take over from the
primary PE. 
And the rest of the domain is still full mesh. 
So it's like I have my VPLS domain with full mesh of PW and should there be
a requirement for a redundant connection on any of the CE sites I'll add the
redundant PE to the domain with the same ve-id and worse metric so that no
one will create PW to it. 
Right?

I actually start to like this :)

Though one thing bugs me still. 
Though this works perfectly when the primary PE fails - how do you tell BGP
that the CE-PE link is down on the primary PE and that it needs to withdraw
its advertisement or perform a local repair if that's possible. 

adam

-----Original Message-----
From: cisco-nsp [mailto:cisco-nsp-bounces at puck.nether.net] On Behalf Of
Caillin Bathern
Sent: Wednesday, April 24, 2013 4:03 PM
To: cisco-nsp at puck.nether.net
Subject: Re: [c-nsp] BGP Signalled VPLS

>Alright so once the VE ID and offset is the same on two PEs, the remote
PEs will create PW only to one of them. 
>This is taken care of the standard VPLS split-horizon rule though.
Not so - if PE-A and PE-B are not using the same VE ID/offset then they will
create a PW between themselves.  This then creates a loop
CE-A---PE-A---PE-B---CE-A.  No split-horizon involved there.  Similarly if
PE-C has active PW to both PE-A and PE-B then when it sends a BUM frame
across the VPLS, the CE-A will receive the BUM frame twice.

>What about PW between PE-A and PE-B (is that PW not created if these
two have now the same VE ID and offset?
Correct
>So that it solves the looping of BUM traffic originated at CE-A sent to
PE-A sent to PE-B and sent to CE-A. 
>But it affects every other PE in the domain as well. 
>So I'd cripple the whole VPLS domain in order to avoid loop between two
PEs, I guess I don't like that.
How does it cripple the VPLS, what do you mean by that?
It provides you an active-backup link to the CE at that site with no
requirement for xSTP.

Caillin

-----Original Message-----
From: Adam Vitkovsky [mailto:adam.vitkovsky at swan.sk]
Sent: Wednesday, 24 April 2013 11:50 PM
To: Caillin Bathern; cisco-nsp at puck.nether.net
Subject: RE: [c-nsp] BGP Signalled VPLS

Alright so once the VE ID and offset is the same on two PEs, the remote PEs
will create PW only to one of them. 
This is taken care of the standard VPLS split-horizon rule though. 

What about PW between PE-A and PE-B (is that PW not created if these two
have now the same VE ID and offset? 
So that it solves the looping of BUM traffic originated at CE-A sent to PE-A
sent to PE-B and sent to CE-A. 
But it affects every other PE in the domain as well. 
So I'd cripple the whole VPLS domain in order to avoid loop between two PEs,
I guess I don't like that. 

I'd rather relay on MC-LAG or REP or MST access gateway in order to avoid
loops. 

adam
-----Original Message-----
From: cisco-nsp [mailto:cisco-nsp-bounces at puck.nether.net] On Behalf Of
Caillin Bathern
Sent: Wednesday, April 24, 2013 2:54 PM
To: cisco-nsp at puck.nether.net
Subject: Re: [c-nsp] BGP Signalled VPLS

Aaron,

I think bep is referring to EVPN in his comment.  In BGP signalled + BGP AD
VPLS, you don't use a specific split-horizon label.  Instead if you think of
the most simple case where you have three PEs: PE-A, PE-B and PE-C.  PE-A
(better local pref)and PE-B (worse local pref) are multi-homed to site CE-A
and PE-C is single-homed to site CE-C.
Now if you consider the BGP advertisements received at PE-C, it has two
possible routes to site CE-A (distinguished by the identical RD, VE ID and
VE block offset).  Hence PE-C selects one of these two sites as its route to
CE-A via BGP route selection process, being via PE-A because of local-pref.
Now PE-C sets up a single BGP signalled pseudo-wire between PE-C to PE-A and
begins forwarding traffic.  Any traffic received at PE-B is dropped because
it has no pseudo-wires set up to any other PEs and there is no loop in the
network.
J* have a good article describing exactly that here
http://www.juniper.net/techpubs/en_US/junos/information-products/topic-c
ollections/nce/bgp-vpls-multihoming/validating-a-bgp-based-vpls-multihom
ing-configuration.pdf

EVPN is a little different because in that above example, PE-C would have
set up pseudo-wires to both PE-A and PE-B allowing active-active forwarding.
Any BUM traffic will still be forwarded using split-horizon though using the
split-horizon label.  At least that is my understanding..

Caillin

-----Original Message-----
From: Aaron [mailto:aaron1 at gvtc.com]
Sent: Wednesday, 24 April 2013 3:42 AM
To: bep at whack.org
Cc: Caillin Bathern; cisco-nsp at puck.nether.net; adam.vitkovsky at swan.sk;
'Saku Ytti'
Subject: RE: [c-nsp] BGP Signalled VPLS

Well why didn't you just say so!  Haha, Seriously, thanks a bunch bep.
I'll begin reading up on this.

Aaron

-----Original Message-----
From: Bruce Pinsky [mailto:bep at whack.org]
Sent: Tuesday, April 23, 2013 12:19 PM
To: Aaron
Cc: 'Caillin Bathern'; cisco-nsp at puck.nether.net; adam.vitkovsky at swan.sk;
Saku Ytti
Subject: Re: [c-nsp] BGP Signalled VPLS

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

Aaron wrote:
> Thanks Caillin/Saku/Adam, this differentiation of VPLS LDP Sig 
> compared to BGP Sig as it relates to loop prevention during redundant 
> pe/ce at edge is of interest to me...(I actually had a l2 forwarding 
> loop scare me to death and had to shut down backside c-to-c during 
> maintenance window a few months ago)....i walked away from that with a

> big question in my head as to how does customer spanning tree feed 
> into the loop prevention of split horizon groups within a vpls as how 
> pw forwarding treatment occurs...and I thought to myself , it probably

> doesn't... which has had me wondering about this for a few months....
> 
> BUT, now y'all mention that bgp signaled vpls as it relates to 
> redundant pe/ce avoids this.... correct?  I have adam's link, thanks 
> adam, but does anyone have more links related to understanding all
that?
> 
> Adam's implementation link....
> 
> http://www.cisco.com/en/US/docs/routers/asr9000/software/asr9k_r4.3/lx
> vpn/co
> nfiguration/guide/lesc43pbb.html#wp1183684
> 

By the selection of a Designated Forwarder via the Ethernet AD route
advertisement and the use of a split-horizon label appended to
multi-destination packets.

- --
=========
bep

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.17 (MingW32)
Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/

iEYEARECAAYFAlF2wpMACgkQE1XcgMgrtybrjQCePMQp1veqynrm8qcWlfqcz325
vZcAoJRF1aOQb7Iz/1qsMN4fyMRm+G9u
=IsdS
-----END PGP SIGNATURE-----

--
Message  protected by MailGuard: e-mail anti-virus, anti-spam and content
filtering.http://www.mailguard.com.au/mg

_______________________________________________
cisco-nsp mailing list  cisco-nsp at puck.nether.net
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/

_______________________________________________
cisco-nsp mailing list  cisco-nsp at puck.nether.net
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/