[j-nsp] layer 2 sp services
Alexander Marhold
alexander.marhold at gmx.at
Thu Aug 3 15:00:03 EDT 2017
>Can't ECMP L2 ?
> I thought I read otherwise... ?
O yes you are right,
EVPN can do L2 ECMP , the method is called "Aliasing" and by adding the ESI
to the mac type 2 announcement, other nodes can send also to second PE on an
all-active multihomed segment using same BGP-EVPN label but other next hop.
Only layer 3 all active loadbalancing does not work now, it was implemented
but silently removed by Juniper due to problem
All above is regarding EVPN over MPLS, EVPN over VXLAN is a little bit
different, VP-LAG ( L3-all active) was mentioned in a slide by d.hanks but I
do not know if already implemented, L2 all-active should work ( have not
tested it on QFX now)
Regards
Alexander Marhold
-----Ursprüngliche Nachricht-----
Von: juniper-nsp [mailto:juniper-nsp-bounces at puck.nether.net] Im Auftrag von
Aaron Gould
Gesendet: Donnerstag, 3. August 2017 20:36
An: adamv0025 at netconsultings.com; juniper-nsp at puck.nether.net
Betreff: Re: [j-nsp] layer 2 sp services
Thanks Adam....
Can't ECMP L2 ?
I thought I read otherwise... ?
https://www.nanog.org/sites/default/files/monday_general_hankins_vpn_2.pdf
slide 8 of 34 ...last bullet "Enables traffic load balancing for multihomed
CEs with ECMP MAC routes"
-Aaron
-----Original Message-----
From: adamv0025 at netconsultings.com [mailto:adamv0025 at netconsultings.com]
Sent: Sunday, July 30, 2017 4:07 AM
To: 'Aaron Gould' <aaron1 at gvtc.com>; juniper-nsp at puck.nether.net
Subject: RE: [j-nsp] layer 2 sp services
> Of Aaron Gould
> Sent: Friday, July 28, 2017 1:09 PM
>
>
>
> My gosh, the layer 2 service provider topic is pretty extensive. I'm
trying to
> learn about all of these.
>
Yeah there was a 2nd renaissance of L2 services demand around 5 years ago
due to all the cloud and DCI hype which resulted in a lot of new standards
especially around EVPN, all the rest are now considered legacy.
Oh, btw the VXLAN doesn't belong with the bunch, That one belongs to the
Ethernet in Ethernet/IP overlay for DC fabric together with TRILL,
Fabric-Path and PBB, happy reading :)
And after all these studies comparing various redundancy options of BGP CP
for EVPN and PBB-EVPN and VXLAN GW, etc..., you'll find out that you simply
can't ECMP L2 as you would a L3 traffic, cause you simply can't associate
single MAC address with two ports, how stupid is that right?
But definitely worth the adventure.
adam
netconsultings.com
::carrier-class solutions for the telecommunications industry::
>
>
> I think I have rfc's and terminology correct. but maybe not. correct
> me
where
> needed.
>
>
>
> - BGP L2VPN VPWS RFC 6624 BGP AD BGP SIG PP draft-kompella
>
> - BGP L2VPN VPWS RFC 6074 BGP AD LDP SIG PP FEC 129
>
> - BGP L2VPN VPLS RFC 4761 BGP AD BGP SIG MP
>
> - BGP L2VPN VPLS RFC 4762 BGP AD LDP SIG MP
>
> - Martini RFC 4447 no AD LDP SIG PP
>
> - EVPN - don't know much here yet
>
> - VXLAN - don't know much here yet
>
>
>
> I'm pretty sure I use Martini/rfc4447 (l2circuits) and BGP/LDP vpls
> rfc
4762
> extensively in my network (routing-instance . instance-type vpls
> rd/rt, etc).but some of these other ones I don't think I'd ever really
> use. like
this
> one that I'm currently learning about. rfc6624/kompella. nope, I don't
think I
> would use this one. Do y'all use kompella ?
>
>
>
> About kompella/rfc6624, I want to make sure I understand the NLRI
> thing as far as knowing how to compute the incoming label and outgoing
> label for a certain p-to-p pw to a remote site.
>
>
>
> I this correct ? ----> label base local + remote-site-id -
> label-offset-local = label value
>
>
>
> Maybe it's like this.
>
> 800006+15-15=800006 (in label)
>
> 800002+23-23=800002 (out label)
>
>
>
> I need to know how did the Local Label Base get calculated ? where
> did
> 800002 and 800006 come from ?
>
>
>
> Also, does the Offset always equal the remote site id ?
>
>
>
> .lab output.
>
>
----------------------------------------------------------------------------
> ----
>
> Instance: vpn-a
>
> Edge protection: Not-Primary
>
> Local site: CE-A (23)
>
> Number of local interfaces: 1
>
> Number of local interfaces up: 1
>
> lt-0/1/0.11 15
>
> Label-base Offset Size Range Preference
>
> 800006 15 2 1 100
>
> status-vector: 0
>
> connection-site Type St Time last up # Up
trans
>
> 15 rmt Up Jul 27 23:53:36 2017
1
>
> Remote PE: 1.1.255.1, Negotiated control-word: Yes (Null)
>
> Incoming label: 800006, Outgoing label: 800002
>
> Local interface: lt-0/1/0.11, Status: Up, Encapsulation: VLAN
>
> Flow Label Transmit: No, Flow Label Receive: No
>
> Connection History:
>
> Jul 27 23:53:36 2017 status update timer
>
> Jul 27 23:53:36 2017 PE route changed
>
> Jul 27 23:53:36 2017 Out lbl Update 800002
>
> Jul 27 23:53:36 2017 In lbl Update 800006
>
> Jul 27 23:53:36 2017 loc intf up lt-0/1/0.11
>
>
----------------------------------------------------------------------------
> ----
>
> Instance: vpn-a
>
> Edge protection: Not-Primary
>
> Local site: CE-A (15)
>
> Number of local interfaces: 1
>
> Number of local interfaces up: 1
>
> lt-0/1/0.2 23
>
> Label-base Offset Size Range Preference
>
> 800002 23 2 1 100
>
> status-vector: 0
>
> connection-site Type St Time last up # Up
trans
>
> 23 rmt Up Jul 27 23:53:36 2017
1
>
> Remote PE: 1.1.255.4, Negotiated control-word: Yes (Null)
>
> Incoming label: 800002, Outgoing label: 800006
>
> Local interface: lt-0/1/0.2, Status: Up, Encapsulation: VLAN
>
> Flow Label Transmit: No, Flow Label Receive: No
>
> Connection History:
>
> Jul 27 23:53:36 2017 status update timer
>
> Jul 27 23:53:36 2017 PE route changed
>
> Jul 27 23:53:36 2017 Out lbl Update 800006
>
> Jul 27 23:53:36 2017 In lbl Update 800002
>
> Jul 27 23:53:36 2017 loc intf up lt-0/1/0.2
>
>
>
>
>
> - Aaron Gould
>
>
>
> _______________________________________________
> juniper-nsp mailing list juniper-nsp at puck.nether.net
> https://puck.nether.net/mailman/listinfo/juniper-nsp
_______________________________________________
juniper-nsp mailing list juniper-nsp at puck.nether.net
https://puck.nether.net/mailman/listinfo/juniper-nsp
More information about the juniper-nsp
mailing list