RE: : Question on VRFs and route targets

: FutureNet 2008 Program Now Available

Re: : Question on VRFs and route targets

Hi Scott,

You could simply define a type 'central service' VPN like

vrf_hub_a
  rd 1:1
  rt_export 1:1
  rt_import 1:2

vrf_spoke_b
  rd 1:2
  rt_export 1:2
  rt_import 1:1

vrf_spoke_c
  rd 1:3
  rt_export 1:2
  rt_import 1:1

It is important that the 2 spoke sites uses different vrf's (different RDs) if connected to the same PE otherwise
the two spokes would have direct connectivity!
You may use the same export/impoprt values for the spoke sites as these anyways don't import prefixes
from each other.
You may consider, that a default route exported e.g. by hub_a may allow packets to be exchanged between
the spokes! 

If you want packet between spoke_b and spoke_c to be routed by the hub_a customer site, you may need 2 interfaces
on PE1, one for receiving packets and the other for sending, then the CE may decide how/if to route packets between
the spokes.

-Frederik

On Friday 28 December 2007, Scott Chiefs wrote:
> I have a question for the experts on this list regarding the use of route
> targets and connectivity in a L3VPN.  Here's the scenario:
> 
> A provider has a customer that is looking for a layer 3 MPLS/BGP VPN between
> 3 of their sites (a,b,c).  However, the customer does not want full
> connectivity between their 3 sites from the provider.  They'd like it set up
> in a hub and spoke model, where A is the hub and B,C are the spokes.  So,
> they do not want traffic from B and C to be routed directly to one another
> through the provider cloud.
> 
> How does the provider build this for the customer in their network?  Do they
> build 2 VPNs --> VPN1 for sites A&B and VPN2 for sites A&C (so PE2 and PE3
> would each have 1 VRF configured and PE1 would have 2 VRFs configured)?  Or,
> do they build one VPN (single VRF on all PEs for this customer) and use the
> route target attribute to accomodate the hub and spoke configuration
> required by this customer?
> 
> Putting my question another way -- within a L3VPN, is full mesh, customer IP
> routing connectivity (in the provider network), ALWAYS established between
> CE's (assuming serving PEs have single VRF for the customer)? In which case
> --- a provider would have to acccomodate the customer request by building
> seperate VRFs/VPNs for the customer?    Or can route targets be used to
> constrain distribution of customer routes among PEs to accomodate customer
> routing demands like hub and spoke?
> 
> CE(a)
> |
> |
> PE(1)
> |
> P
> |\
> P-P---PE(2)----CE(b)
> |
> |
> PE(3)
> |
> |
> CE(c)
> 
> Any responses are appreciated.  Thanks!
> 

: OSPF & BGP Forum

: Regarding MPLS VPN (VPLS)

Re: : VPLS operations deployment statistics

Re: : VPLS operations deployment statistics

Hi,

Yes you can have a router that functions as a P router and also as a  
PE router.

In answer to your second question, VPLS would be enabled on PE  
devices. It does not really matter if the PE (LER) also is a P (LSR).

Another way to look at it; if you enabled VPLS on a P router, the  
router is now a P/PE device.

There are pro's and con's that are a separate discussion around  
network architecture when you are deciding if you should have a  
collapsed P/PE network or if you require functionally specific  
devices. You would then have to examine some key points such as:  
cost, risk of high touch environments on the core, growth models,  
etc ... Although this is VPLS-agnostic and rather a discussion on  
technology choices in design.

Kind regards,
Truman Boyes

On 25/01/2008, at 1:19 AM, Akshay Balhara wrote:

> Dear Bala,
> Here is some doubts.
>
> I just want to know that Can we use P Routers as a PE Router for  
> VPLS Implementation.
>
> Simplely, Can I implemet VPLS on Core MPLS (On P Routers)
>
> Regards,
> Akshay Balhara
>
>
>
>
>
> On Tue, 18 Dec 2007 Bala Venkata wrote :
> >Hello-
> >
> >Wondering if anyone has real-world (deployment-experienced if  
> possible) based numbers for the following:
> >
> >- Size of a VPLS network
> >- No. of PEs participating in atleast one VPLS
> >- No. of sites in each PE in each VPLS (an average number that is)
> >- No. of LSPs between all PEs.
> >
> >It is possible to get a theoretical maximum for each of course. I  
> am looking for 'typical' numbers or atleast what you observe in  
> your network.?Without offending anyone, if you could include vendor  
> info that'd help as well :)
> >
> >
> >Thanks !
> >
> >/bala
> >
> >
> >_____________________________________________________________________ 
> ___
> >More new features than ever.  Check out the new AIM(R) Mail ! -  
> http://webmail.aim.com
>
>
>

RE: : VPLS operations deployment statistics

Balhara,

Can you tell us the reasoning behind this design? Or you just asking if it's possible. As Truman said,
technically YES it's possible, although not the optimum design.

Thanks,

Liban Mohamed
NTAC-IP
Sprint/Nextel
www.sprint.net
liban.mohamed <at> sprint.com
(W) 678-291-3438
(PCS) 404-441-9701

-----Original Message-----
From: Truman Boyes [mailto:truman <at> suspicious.org]
Sent: Thursday, January 24, 2008 10:36 AM
To: Akshay Balhara
Cc: Bala Venkata; mpls-ops <at> mplsrc.com
Subject: Re: [MPLS-OPS]: VPLS operations deployment statistics

Hi,

Yes you can have a router that functions as a P router and also as a
PE router.

In answer to your second question, VPLS would be enabled on PE
devices. It does not really matter if the PE (LER) also is a P (LSR).

Another way to look at it; if you enabled VPLS on a P router, the
router is now a P/PE device.

There are pro's and con's that are a separate discussion around
network architecture when you are deciding if you should have a
collapsed P/PE network or if you require functionally specific
devices. You would then have to examine some key points such as:
cost, risk of high touch environments on the core, growth models,
etc ... Although this is VPLS-agnostic and rather a discussion on
technology choices in design.

Kind regards,
Truman Boyes

On 25/01/2008, at 1:19 AM, Akshay Balhara wrote:

> Dear Bala,
> Here is some doubts.
>
> I just want to know that Can we use P Routers as a PE Router for
> VPLS Implementation.
>
> Simplely, Can I implemet VPLS on Core MPLS (On P Routers)
>
> Regards,
> Akshay Balhara
>
>
>
>
>
> On Tue, 18 Dec 2007 Bala Venkata wrote :
> >Hello-
> >
> >Wondering if anyone has real-world (deployment-experienced if
> possible) based numbers for the following:
> >
> >- Size of a VPLS network
> >- No. of PEs participating in atleast one VPLS
> >- No. of sites in each PE in each VPLS (an average number that is)
> >- No. of LSPs between all PEs.
> >
> >It is possible to get a theoretical maximum for each of course. I
> am looking for 'typical' numbers or atleast what you observe in
> your network.?Without offending anyone, if you could include vendor
> info that'd help as well :)
> >
> >
> >Thanks !
> >
> >/bala
> >
> >
> >_____________________________________________________________________
> ___
> >More new features than ever.  Check out the new AIM(R) Mail ! -
> http://webmail.aim.com
>
>
>

-------
The MPLS-OPS Mailing List
Subscribe/Unsubscribe:  http://www.mplsrc.com/mplsops.shtml
Archive: http://www.mplsrc.com/mpls-ops_archive.shtml

Re: : OSPF & BGP Forum

: Load Balancing in the MPLS Core