TE for VPLS customers

Unanswered Question
Feb 15th, 2010
User Badges:

Hi all,

Is it possible to configure TE for individual customer, instead to forward all traffic (all customers traffic) in single tunnel. where all customers are in VPLS (l2VPN).

eg. A diagram is attached herewith for further reference : one TE tunnel for customer CE-a and same like for others customers.

Thank you.

Uttam

Attachment: 
  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 4 (1 ratings)
Loading.
Laurent Aubert Mon, 02/15/2010 - 19:49
User Badges:
  • Cisco Employee,

Hi,


With VPLS, you can't map the attachment circuit to a TE tunnel. You can only map a TE tunnel to a VPLS neighbor via a pseudowire-class.


HTH


Laurent.

urs.shrestharana Mon, 02/15/2010 - 22:09
User Badges:

Hi,



I think this will not help my problem, is there any other way to split the tunnel, my requirement is to carry half of the customers via tun1 and half of the customer via tun2.

thanks,

Uttam

Laurent Aubert Tue, 02/16/2010 - 08:42
User Badges:
  • Cisco Employee,

Hi,


Are all the CE routers sharing the same subnet ? If not you don't need VPLS and EoMPLS is enough. With EoMPLS you could have a seperate TE tunnel per CE.


Thanks


Laurent.

urs.shrestharana Tue, 02/16/2010 - 22:23
User Badges:

Hi Laurent

Yes, there are many customers they sharing the same IP subnet, and the only requirement is VPLS. I think use one TE Tunnel per customer it’s a waste of resources and not a scalable design. now let’s categorise the customers in two groups weighted and light, we will have two links from each PE to PE say primary and backup and of course primary will have higher BW path then the Backup. now what we have to do is, if primary link goes down, only the customers of group weighted can use or be allowed on the backup path. Group light customer will not be allowed to use the backup path in such case.

Please help me with the scenario, how can it be achieved.

Thank you,

Uttam




Laurent Aubert Wed, 02/17/2010 - 09:06
User Badges:
  • Cisco Employee,

Hi,


You could use H-VPLS to achieve what you want. To do that you need a 3rd PE where you end into a VFI pt-2-pt PW from the PE connected to the CE. This way you could map each pt-2-pt PW to a unique TE tunnel.


HTH


Laurent.

urs.shrestharana Wed, 02/17/2010 - 22:56
User Badges:

Hi Laurent,

Really appreciate for  your regular help.

I am not able to understand your last post the H-VPLS concept and that the 3rd PE, is it a u-PE where CEs are connect, and Q-in-Q is used to connect n-PEs. May be I am not good on how to use TE to solve my problem. Could you please provide me the related docs or your small example will really helpful for me.

Thanks,

Uttam

Laurent Aubert Fri, 02/19/2010 - 14:00
User Badges:
  • Cisco Employee,

The idea I have in mind is:


CE-a ----- PE1---------

                               MPLS CLOUD  -------PE3


CE-b------PE4----------


You do EoMPLS between PE1 and PE3 and between PE4 and PE3. You can map each PW to a different TE tunnel and you do VPLS on PE3 only.


I must admit it's not an ideal solution.


Are CE-a,b and c the same customer ? If yes I don't understand why only few of them should benefit of a backup path ?


Thanks


Laurent.

urs.shrestharana Sat, 02/20/2010 - 01:55
User Badges:

Hi Laurent,

There can be two types of customers one who pay high amount another who don’t. So the backup path will be provided only for those customers who pay high. Regarding customers a, b and c they can be the same customers, if they are sharing inter branch connectivity, but there can be so many customers connected to PE1 to reach behind the PE3.

I don’t think EoMPLS PW class to map different TE tunnel (Tunnel between PE1 to PE3) is a bad idea, if it is really works!

Thank you,

Uttam

Laurent Aubert Mon, 03/01/2010 - 20:06
User Badges:
  • Cisco Employee,

Hi,


If you have two customers and each one has two CE's one connected to PE1 and the other one connected to PE3, you can use EoMPLS and bind each PW to a different TE tunnel.


If each of your customer has more than two CE's, then you will have a bridge domain per customer and so you will have different PW's even between the same PE's and can bind each session to a different TE tunnel as well.


What is not working is when you are within the same bridge domain. In this case several CEs attached to the same two PE's will not be able to use different TE tunnels.


HTH


Laurent.

Chetan Kumar Ress Mon, 03/01/2010 - 21:54
User Badges:
  • Silver, 250 points or more

Hi Laurent


Thanks . Great stuff provided by You .


I Just need to understand the Difference between VPLS & H-VPLS .


And clarify whether i use VPLS or H-VPLS the core will be used only IP/MPLS . The packet will be encap with IGP & MPLS Label to forward the packet to neighbore PE.



And please clarify the difference between MPLS & EOMPLS. As per me i understand that in MPLS it work with IGP core and EoMPLS work with Q-n-Q & PW. If possible please give me nay link where i can get the same example.

Laurent Aubert Tue, 03/02/2010 - 07:37
User Badges:
  • Cisco Employee,

Hi,


Here is a link about VPLS and H-VPLS. The difference between the two is H-VPLS scale better by adding an edge access  build on Q-in-Q or EoMPLS network.


http://www.cisco.com/en/US/tech/tk436/tk891/technologies_white_paper09186a00801f6084.shtml


EoMPLS is using LDP to build pt-2-pt PW so you need a MPLS aware backbone to run this service.


http://www.cisco.com/en/US/docs/ios/mpls/configuration/guide/mp_any_transport_ps10591_TSD_Products_Configuration_Guide_Chapter.html#wp1047047


HTH


Laurent.

judebryant Fri, 03/12/2010 - 07:17
User Badges:

Hello to all,


is there a reason you need VPLS.  do you need point to multipoint.


how about using pseudowires to provide the individual tunnels.


create 2 service instances on pe3 which will make deciscions based on vlan or vlans.  This decision will direct traffic into the correct tunnel to the opposite router.


Hope I followed this thread correctly and provided some useful information


Regards

Jude Bryant

Pioneer Telephone...

judebryant Mon, 03/15/2010 - 07:34
User Badges:

Hello Group,


Hello Shivlu Jain


I like the two articles you have presented, however in a service provider environment why would you have all of this layer3 routing going on.  I feel the only time you would need this if sharing routes with a customer.


Please explain to me


In my environment I have an mpls cloud using isis as the igp.  There is no BGP.  The only advantaged of BGP in a non route sharing environment if for auto-vpls.


or I could be wrong.


The second article intrigues me.  I have never read anything about these two modes of pseudowires.  Also where does spanning tree come into play with pseudowires.  I would think spanning tree is not at all present in your mpls cloud.


as for traffic engineering, how about using the commands to create pseudowires based on ldp.  As in forcing the pseudowire path to a predetermined path instead of the igp making the decisions.  Then your backup pseudowires can be set for an alternate route of that alternate route.


just curious


Regards

Jude Bryant

Pioneer Telephone

judebryant Mon, 03/15/2010 - 09:00
User Badges:

Please not everyone.


When using pseudowires, customer mac address's are not learned by your core mpls network.


however, in vpls, customer mac addresses are learned.  So be careful when using vpls, unless you have implemented 802.1ah



Regards

Jude Bryant

Pioneer Telephone

judebryant Mon, 03/15/2010 - 14:11
User Badges:

OK,


so I did some research.



example

Ethernet pseudowire type/mode tagged or 0x00004


(lan1)==(tagged traffic)==>{interface}PE/P[R1]{mpls cloud}[R2]PE/P{interface}==>(tagged traffic)==(lan2)



vlan 123 is not globally defined.

mac address from both lan1 and lan2 are not learned within the cloud

this is a point to point pseudowire.


router 1

lo0

ip address 10.10.10.1


interface 3/0/1

service instance 100 ethernet

encapsulation dot1q 123

rewrite ingress tag pop 1 symmetric

xconnection 10.10.10.2 100 encapsulation mpls



router 2

lo0

ip address 10.10.10.2


interface 3/0/1

service instance 100 ethernet

encapsulation dot1q 123

rewrite ingress tag pop 1 symmetric

xconnect 10.10.10.2 100 encapsulation mpls


the packets ingress from lan 1 is tagged with vlan123

because of this the service instance is making decisions based on matching this tagged traffic (encapsulation dot1q 123)


example

Ethernet pseudowire type/mode raw or 0x00005


(lan1)==(tagged or untagged traffic)==>{interface}PE/P[R1]{mpls cloud}[R2]PE/P{interface}==>(tagged or untagged traffic)==(lan2)


vlan 123 is not globally defined.

mac address from both lan1 and lan2 are not learned within the cloud

this is a point to point pseudowire.


the service instances do not try to match vlan tags so we use the word default for encapsulation type.  not decisions are made.  this is raw mode

traffic in traffic out


router 1

lo0

ip address 10.10.10.1


interface 3/0/1

service instance 100 ethernet

encapsulation default

rewrite ingress tag pop 1 symmetric

xconnection 10.10.10.2 100 encapsulation mpls



router 2

lo0

ip address 10.10.10.2


interface 3/0/1

service instance 100 ethernet

encapsulation default

rewrite ingress tag pop 1 symmetric

xconnect 10.10.10.2 100 encapsulation mpls


the packets ingress from lan 1 is tagged with vlan123 or untagged

now we can create a service instance that is not making decisions based on matching this tagged traffic (encapsulation default)

anything that comes in on this port tagged or untagged goes thru


this is similar to a layer2 tunnel, however this tunnel can be protected by a backup pseudowire..



Hoped I explained this in a clear matter.  If not please ask any questions you mave have.


Please refer to the document provided by

shivlu jain

Modes Of Ethernet Over MPLS (EoMPLS)


thanks Shivlu


and Regards


Jude Bryant

Pioneer Telephone

Actions

This Discussion