MPLS backbone with L3VPNs over IPv6 only

[__djidji__]

Hello,

I'm looking to build an MPLS backbone and implement L3 VPN services on top of it. However the underlying IP network has to be IPv6 ONLY. From what I can tell the LDPv6 is not really an option as is not actively developed. RSVP-TE may be an option but I'm unclear if it can be used for this purpose and with my limitations - documentation is unclear. I'm looking at roughly 300-400 PEs with number of services / VRFs less than 10. Any suggestions?

Thanks,

George

[Harold Ritter]

Hi George,

There is currently no option to build an MPLS backbone over an IPv6 only core. LDPv6 is supported in IOS-XR, but  L3VPN services are not supported. RSVP-TE is currently not defined for IPv6 only.

On the other hand, you could deploy L3VPN services using Segment Routing V6 (SRv6). SRv6 is relatively new and does not require MPLS at all in order to offer services such as L3VPN. A few customers have already officially announced SRv6 deployments.

Please refer to the following site for more information on SRv6:

https://www.segment-routing.net/tutorials/2017-12-05-srv6-introduction/

Regards,

[__djidji__]

Hi Harold,

Thanks for that. Just for my education, what do you mean by: "RSVP-TE is currently not defined for IPv6 only"? 

As far as SRv6 goes are the segments identifying the service / VRF distributed between PEs using MP-BGP - similar to MPLS labels with VPNV4?

Thanks again,

George

[Harold Ritter · Accepted solution]

> Just for my education, what do you mean by: "RSVP-TE is currently not defined for IPv6 only"? 

I mean, defined by IETF to operate in an IPv6 only network like LDP has for instance (RFC7552).

> As far as SRv6 goes are the segments identifying the service / VRF distributed between PEs using MP-BGP - similar to MPLS labels with VPNV4?

Yes, SRv6 uses MP-BGP (VPNv4) exactly like with MPLS VPN.

Regards,

[tulasi rami reddy]

Hi Harold,

Do we support L3VPN over MPLS backbone with IPv6 only core currently?

In dual-stack deployment with both IPv4 & IPv6 (for SRv6) MPLS based L3VPN advertisement,

What is the preference for selection of best path?

Do we need to use some route-maps to change the BGP parameters to prefer one or any other configs?

[Harold Ritter]

Hi,

> Do we support L3VPN over MPLS backbone with IPv6 only core currently?

It is not currently possible to do that. SRv6 is probably the way to go for that.

> What is the preference for selection of best path?

The preference is the same as it would in a standard L3VPN environment.

> Do we need to use some route-maps to change the BGP parameters to prefer one or any other configs?

The same technique as you would currently use with L3VPN applies.

Regards,

[tulasi rami reddy]

Thanks Harold for the clarification. Let me further clarify my question here related to route selection:

ce--ingress PE ----- (MPLS V4 & V6 cloud) ----- egress PE --ce

If ingress PE receives two L3VPN routes, one over IPv4 session(for legacy MPLS) and other over IPv6 session (SRv6), then all BGP parameters for L3VPN would typically be same except the nexthop-ip. In this case the regular BGP bestpath may select the one with lower peering IP. This may not be one interested at ingress..

In this case the selection should be based on some local policy as highlighted in below draft: 

Could you please share if any CLI option for local decision  to SRv6 or what are the migration steps operator expected?

https://tools.ietf.org/html/draft-dawra-bess-srv6-services-00#page-16

5. Migration from MPLS based Segment Routing to SRv6 Segment Routing

Migration from IPv4 to IPv6 is independent of SRv6 BGP endpoints, and
the selection of which route to use (received via the IPv4 or the
IPv6 session) is a local configurable decision of the ingress PE, and
is outside the scope of this document.

Thanks,

Tulasi.

[Harold Ritter]

Hi Tulasi,

Oh, I see. You meant if you receive the same prefix via a given PE, over both MPLS (ipv4) and SRv6 and you want to prefer SRv6, correct?

I see that the draft suggests the following:

Selection of either MPLS encapsulation or SRv6 encapsulation is defined
  by local BGP policy.

I do not think this is actually available yet, but the same could be accomplished via the setting of local preference either on the egress or ingress PE. Having the local preference set higher on prefixes learnt via SRv6 would have the effect of the ingress PE selecting these prefixes.

Regards,

[tulasi rami reddy]

Yes, this is the scenario I was looking.

Thanks Harold for confirmation.

BTW, Do you see any traction for L3VPN over MPLS backbone with IPv6 only core in near future or SRv6 approach is preferred option for IPv6 core migration?

Thanks,

Tulasi.

[Harold Ritter]

I do not see a big demand for service providers wanting to deploy IPv6 only MPLS core for now. Those going this way are adopting SRv6. 

Regards,