Hi,
You have posted the same question in the LAN Routing & Switching forum:
https://supportforums.cisco.com/discussion/12554041/mpls-concepts-and-queries
Please delete it first so that the answers do not get spread over two threads.
1. Can we configure VRF without route distinguisher ?
Technically, we can, but it would cause major problems with BGP. Older IOSes allow omitting RD from a VRF definition, newer IOSes require you to configure the RD before using it.
The RD is used in BGP to differentiate between same-addressed networks located in different VRFs. Without RD, network 10.0.0.0/8 in VRF A would be indistinguishable from network 10.0.0.0/8 in VRF B to BGP. If BGP advertised networks from both VRFs without an RD then the "second" 10.0.0.0/8 (from whichever VRF) would appear to be an update of the "first" 10.0.0.0/8. The receiving router would learn only about the "second" 10.0.0.0/8 and it would forget about the "first" 10.0.0.0/8.
Because VRFs on a single router are required to have unique RDs, this RD is used to "inflate" the address of each network in that VRF, making it unique. If VRF A used RD 1:1 and VRF B used RD 1:2, then BGP would advertise networks 1:1:10.0.0.0/8 and 1:2:10.0.0.0/8, and these two are clearly different so BGP won't get confused.
The concept of RD is basically just a workaround for a specific BGP's behavior - that if the same network is advertised multiple times, it is treated as an update, and only the most recently received advertisement is used.
2. In MPLS, penultimate hop popping remove label at P router itself .. for that do we enable any specific configuration ?
No, PHP is an automatic feature in MPLS and you do not explicitly configure it.
3. In L2VPN and L3VPN who assign inner and outer labels is that LDP or BGP ?
In both these VPNs, the outer label (also called the transport label) is assigned by LDP, as it represents an internal Label Switched Path toward the far end, either a pseudowire tailend or an egress PE. The internal label in L2VPN is also advertised by LDP in a targeted session towards the pseudowire tailend. In L3VPN, internal label is advertised by BGP.
4. How do we define that we configured MP BGP in MPLS ? if we configure address family does it mean its MPBGP ?
Basically, yes, this is correct. MP-BGP is not a separate BGP version. It is just a set of extensions to its basic message types so that using the good old BGPv4 messages, we can advertise very diverse information elements. If you can configure address families in BGP then it means that your router supports the multiprotocol extensions to BGP. Quite frankly, we run MP-BGP all the time nowadays so it's not even worth mentioning.
Even if I configure address family with Ipv4 unicast under BGP configuration, how we are able to execute the command show VPNV4 unicast <VRF name> ?
I do not fully understand this question. However, the fact that you can run an arbitrary show command is not surprising: You can do show ipv6 route even on a router without configuring IPv6, or you can do show access-lists even if no ACLs are created. However, the output in such case would be empty, or you'll get some error message. It's the same with show bgp vpnv4 unicast command - you can enter it anytime. Whether you get a reasonable output - that's another story. In general, to get anything useful, your BGP would need to be configured with address-family vpnv4 unicast.
I configure address family with Ipv4 multicast how does that behave ?
This address family is used for large-scale multicast deployments in which two different autonomous systems want to share and distribute multicast. In multicast routing, each router must perform a check on the incoming multicast packet to verify whether it comes through an interface that leads back to the multicast source. We call this a Reverse Path Forwarding check, or RPF. Inside a single autonomous system, the information for looking up the RPF interface is contained in common routing tables populated by internal routing protocols (OSPF, IS-IS, EIGRP, RIP). However, these protocols do not run between autonomous systems. Therefore, BGP was enhanced with a separate address family that allows one autonomous system to advertise networks in which multicast sources reside to another autonomous system so that it can perform its own independent RPF checks toward these sources. This address family stands separate because the routing information contained there is not used to learn about networks in a different autonomous system and route traffic towards them, but only to perform RPF checks when receiving multicasts from them.
Best regards,
Peter
