Nilesh,
Basic MPLS does not need BGP. It uses standard internal routing protocols to find destinations and paths towards them, and it uses the Label Distribution Protocol to assign and disseminate label bindings to these destinations.
BGP is used in more advanced applications of MPLS, in particular:
- Layer3 VPNs
- Layer2 VPNs (here, also LDP may be used as well)
- VPLS
- 6PE
- 6VPE
- Multicast VPNs
BGP is used in these scenarios because it allows to communicate both networks and corresponding label bindings in a single update, and it allows the exchange of this information between a particular selected pairs of routers - usually the Provide Edge (PE) routers. Other routers do not usually need this information, and the use of BGP allows us to spare them the unnecessary details - a great asset for scalability.
For example, in Layer3 VPNs, the BGP carries the networks of individual customers that are connected to a PE, and conveys this information to another PE where another locations of the same customer may be connected. In addition, it adds an MPLS label to each of these communicated customer networks. However, the internal routers (called P routers) that lie on the path between these PE routers do not see this BGP communication - they do not need BGP, and they do not need to. When a PE routes a packet from a customer to its another location connected at the another PE, the following occurs:
- The local PE looks up the destination network, and finds the label remote PE assigned to this network. These networks and labels were advertised by BGP from the remote PE to the local PE. It adds the first MPLS label to the packet.
- Then, the local PE will try to look up the next-hop for this destination network. This next-hop will be set to the remote PE's address itself. The remote PE's address (usually a loopback) is advertised via the usual IGP, say, OSPF, and the label to it is known via LDP. So the local PE will add another label (a second one) to the packet and send it along the way towards the remote PE.
- All intermediate P routers will look only at the topmost label that is set in such a way to indicate that the packet should arrive at the remote PE. They do not inspect the inner labels, nor do they inspect the destination IP address in the packet. As a result, they do not need to know anything about the customers or their networks.
- The last P router before the PE router removes the topmost label (Penultimate Hop Popping) and the packet will now arrive with only the first label to the remote PE.
- The only label that remained in the packet is the very label that the remote PE assigned to the customer network and communicated to the local PE via BGP. As a result, the remote PE exactly knows which customer network is this particular packet destined for, so it removes the label and forwards the packet to its correct destination.
I believe that if you try to look for words like "mpls fundamentals", "mpls basics", "mpls and bgp" in google you will certainly find lots and lots of useful information. And of course, you are welcome to come here and ask!
Best regards,
Peter
