MPLS noob question
I am just starting to learn about MPLS and I am having a hard time finding information about MPLS from a customers perspective. I understand that there is no specific MPLS configuration on the customers end and that they usually form a BGP neighborship with the PE router and advertise their routes.
So do all of the customers routes get advertised to each other adding all the advertised routes into the customers routing table?
What if the customer has thousands of sites and most sites don't need to communicate with each other? Wouldn't that be a waste of resources on the customers routers with such a large routing table?
What if the customer wants to implement a hub and spoke topology?
If anyone knows of a good read that could answer these questions without going to deep into the service provider end of things I would appreciate it, thanks.
So do all of the customers routes get advertised to each other adding all the advertised routes into the customers routing table?
What if the customer has thousands of sites and most sites don't need to communicate with each other? Wouldn't that be a waste of resources on the customers routers with such a large routing table?
What if the customer wants to implement a hub and spoke topology?
If anyone knows of a good read that could answer these questions without going to deep into the service provider end of things I would appreciate it, thanks.
Comments
Depends, but most of the time: Yes.
What if the customer has thousands of sites and most sites don't need to communicate with each other? Wouldn't that be a waste of resources on the customers routers with such a large routing table?
Routers connected to an MPLS should be able to handle a large BGP routing table. If they only want certain routes available on a site they can configure this on the site's router in the BGP configuration.
What if the customer wants to implement a hub and spoke topology?
The whole point of an MPLS is that you don't need a hub & spoke. You don't have to rely on one site to get to another, they will communicate directly. But if you really wanted to do that, you could do it with manual configuration of the routes on the routers.
However, if you can find it, have a look at the old CCNP R&S ISCW book (from 200
So if I understand correctly take the following example
-you have one main site and thousands of remote locations all conneted by MPLS
-they are all under a 10.x.x.x /24 network.
-The remote sites very rarely have to communicate with each other, but frequently with the main site
***Would this be the best way to achieve the best routing solution via MPLS by not having your remote routers filled with thousands of routes it will never use.
1) At the main site have a heavy duty router taking in all the remote sites routes into its routing table and advertise out a 10.0.0.0/8 summary route.
2) The remote sites would advertise their routes outward but block every incoming advertisement except the summarized route from the main site. In the off chance that a remote site to remote site should need to communicate they can still go thru the main sites router via the summary route?
Honestly though, a thousand routes isn't that big of a deal for a modern router to churn through anyway.
Think about a large Voice network. Once the call setup is performed on internal calls, the the RTP traffic is p2p between the phones. Imagine sending all that RTP back to a core site just for a couple of phones to send packets from one to the other.
One reason why MPLS is so popular with converged networks. End to end QoS being the other big factor.
If you want an intro to route targets (assuming some BGP knowledge here) this guy does a pretty good job of how prefixes can be imported and exported with VRFs and BGP - Route Distinguishers and Route Targets - Packet Life
Remembering that BGP does most of the work here, MPLS just provides the switching mechanism inside the SP between the PE routers so that the internal routers (the P routers) only need to maintain routing information within the SP network rather than having to maintain Internet size routing tables.
Very details-scare overview there but HTH
Your forgetting about the the vrf label that is exchanged via bgp. A label is assigned to each local VRF and distributed to the peer PE using BGP, another label is exchanged for all prefixes in the RIB using LDP which is what you are referring too. When a packet arrives on a local VRF port that is destined for the remote VRF it will be labelled with the remote PE's vrf label and also the next-hop label, hence its a double label stack. Without the inner label the remote router wouldn't know which VRF the packet belongs to.
Bachelors of Science in Telecommunications - Mt. Sierra College
Masters of Networking and Communications Management, Focus in Wireless - Keller