7. What is “Label Switching”?
• Traditionally each router looks “inside” the
packet to determine it’s destination
• Opening the envelope at every sorting office
to work out where it needs to go = SLOW!
8. What is “Label Switching?”
• In MPLS routers have pre-defined label-
switched paths in their routing table
• They already know the best route to varying
destinations
9. What is “Label Switching?”
• Router 1 (PE) adds a shipping label to the
packet
10. What is “Label Switching?”
• The rest of the routers handle the packet with
this label according to their LSP
11. What is “Label Switching?”
• The final router pops the label and delivers as
normal
15. Really quick other stuff – MPLS VPN
• Switch in the cloud (L2 - VPLS)
– Offers a long Ethernet cable
– VPN spanning between sites
– Time critical applications (VoIP, Video)
• L3 VPNs (VPRN)
– VRFs for each customer
– CE and PE peer and exchange routes
– Complex network layout
– Needs MP-BGP
MPLS – Multi Protocol Label Switching is a protocol that service providers run within the core network to speed up routing and allow QoS more easily.
MPLS doesn’t happily fit in either the Data-Link layer, nor the Network layer. It provides enhanced L2 features, but isn’t quite at L3. This is why we say it’s at layer 2.5. It was originally designed to speed up the routing of packets across a service providers network (cloud) by altering the way that the routers examine the packets and make forwarding decisions. However with the development of switching and routing technology (ASICs) these days it’s main use is to provide VPNs and other traffic engineering.
There are many benefits to using MPLS. Lets look at interface independence and scalability.
Before MPLS we would usually connect sites together using some sort of WAN link like Frame Relay circuits. Let’s say we have a site in London, and another in Sheffield. Not too bad – just two sites to connect to each other. But what happens if we try and scale this model? Let’s add Leeds, Edinburgh and Glasgow. Now we have all our sites connecting to each other. What happens if we lose one of our links? Say between Sheffield and Leeds. That means that Leeds, Glasgow and Edinburgh are all offline until the link is restored.
Lets take the same 5 sites and connect them to an MPLS network instead. Now none of the sites is dependent upon another for it’s connection and instead we have a fully meshed one to many relationship. As the service provider gives you virtual circuits between each of your sites each can keep working independently of one another.
Taking our same mesh – MPLS adds another benefit: Interface independence. Lets say that our Leeds site only has 5 employees based there. We don’t want to be shelling out for a 100Mbps Ethernet circuit when a simple ADSL connection will do. Well, MPLS allows you to do this.
In traditional IP networks each router makes a forwarding decision based upon the L3 destination header of each packet. It has to look at every packet in detail! Our poor router feels like a washing machine, constantly rinsing and repeating – not only that but doing this slows it down. It’s like you are sending a letter from New York to London and every sorting office had to open the envelope, look at the address, repackage the letter and send it on to the next sorting office. BORING! Oh, and slow. MPLS gets around this problem by enclosing your letter in an envelope with the destination on the front.
Instead of having to examine the packet, each router now just looks at the label and sends the packet on it’s way according to the pre-defined routes (Label Switched Paths) in it’s VRF. This operation is similar in both logic and speed as to how L2 switches make forwarding decisions. This has the net result of speeding up the whole packet transit.
The router at the “edge” of the service providers network does the label popping, and imposition. So traffic coming into this router will be examined and given a label; whereas traffic exiting at this router will have the label “popped” off, and forwarded as a normal IP packet.
The routers in the middle already know the best route to the destination label, so all they need to do is look at the envelope and read it’s address, then forward it out of the correct interface on it’s onward journey.
And as already mentioned, the PE at the other end pops off the label and delivers the packet as normal.
Looking at the mesh we used earlier we’ve got all these sites connected via their own technology. Each of these technologies runs a different protocol over the top. MPLS is able to handle this without an issue, whereas older technologies would have to run the same protocol throughout the network.
The “P” Routers are at the core of the MPLS network infrastructure. Also called LSR’s = Label Switching Routers.
The “PE” routers are the provider edge routers. So called, because they sit at the edge of the provider’s network. It’s important to remember that there are many PE’s, but you will typically only interface with one.
Last up we have the CE router. CE is Customer Edge, so called because they sit at the edge of your network as the customer. Typically the peering between the PE and CE is done via a routing protocol – most popular is BGP (Border Gateway Protocol). The customers’ network administrator will program their router with an IP address facing the PE. This is usually a /30 for the purposes of the link.
VPLS = Virtual Private LAN Service and allows you to run Layer 2 across multiple sites, and extend your VLAN tagging across the cloud to your remote site.
VPRN = Virtual Private Routed Network is the Layer 3 version of VPLS. This allows SP’s to use the same physical P router for many customers, by segregating their traffic into different VRF’s.
MP-BGP = Multiprotocol BGP