ASR9K VPLS configuration assistance

[phamthecong]

Hi All,

we have 2 ASR9010 IOS XR version 6.0 at 2  sites and would like to span 2 VLANs across the MPLS network using VPLS. The config is below

PE1 

interface bundle-eth2.200  l2transport

encapsulation dot1q 200 , 300

l2vpn

bridge group 2

bridge-domain Spansites

interface bundle-eth2.200

vfi 1

neighbor 10.1.1.1 pw-id 200

==============

PE2 config

nterface bundle-eth2.200  l2transport

encapsulation dot1q 200 , 300

l2vpn

bridge group 2

bridge-domain Spansites

interface bundle-eth2.200

vfi 1

neighbor 10.1.2.1 pw-id 200

=======

Both CEs are running etherchannel and trunk.

will the config above work? will the ASR forward incoming frame from VLAN 200(or 300)  to PW with tag 200(or 300)  or untag?

Regards,

Tao

[Evgeniy Prichinin]

basic config require

 rewrite ingress tag pop 1 symmetric

on l2transport subinterface, so you config will not work.

[DIEGO ZORRILLA]

I have pass trafic with and without the pop symetric in the l2tranpsort sub-interface, not sure if the PW that you are showing are /32, as far as I know that is a MOST.

[xthuijs]

in principle this config will work but I dont think it is a recommended approach, let me explain.

In your scenario you are preserving the tag, so the single pw can carry both and on the remote side there is no issue either.

since you only have one AC and one PW, you might as well go for a cross connect. this is faster in the fowarding path and also saves mac learning resources, no concern about flooding or unnecessary replication when the received packet is a bcast or mcast.

in Xconn, the receive packet is just encapped and sent over, which is therefore much faster in terms of ucode cycles.

config for that would be:

l2vpn

xconn group BLA

p2p MINE

int bundle-e100.20

neighbor 10.1.2.1 pw-id 100

I would either do port mode, considering you mention trunk on the CE side.

Or I would create specific p2p's for those vlans you want to tunnel across.

But you can also leave them like this of course.

If you create individual p2p's or BD's for each vlan separately you can use the pop tag symmetric, which saves a few bytes vlan overhead on the mpls transport.

regards

xander

[phamthecong]

Thank you everyone for your response.

in the XR platform the tag will not be popped  at ingress interface and the incoming frame with tag will be forwarded to PW. the IOS platform will pop the incoming tag and forward untag frame to PW.

For the Xconnect vs BD, that is for future use as that  leaves room to span these vlans across multiple sites.

For each vlan per PW vs multiple vlan(ambiguous encapsulation) per PW, this in an interesting point. when using ambiguous encap, we have overhead. However, it is simple in term management if we have 100+ VLANs to span. Port mode is an option but it will be an issue if some vlans need to translate while others don't.

Xander, 100% aggree with you in the example above, port base or 2 PWs will be better choice.

Once again, thanks everyone.

The