How Does MPLS result in better efficiency

Isnt mpls supposed to make routing tables smaller?
But mpls ultimately has to make use of the next hop to deduce where to send the labels.
So how does mpls save computational reources?

Find more posts tagged with

Comments

EdTheLad

It will make the routing table of the P routers smaller not the PE. With an mpls core the P routers don't need to run bgp, they only need to know how to get to the next-hop i.e. the egress PE. An IGP is run in the core, all PEs advertise their next-hop addresses into the IGP so every P knows how to reach every PE next-hop. With mpls enabled, every PE will assign a label to its local loopback, these labels will be learned via ldp by all other P and PEs.

PE1 will learn external routes from PE2 via iBGP, PE1 will learn the next-hop of PE2 via IGP, when packets are sent to PE2 they will be impose the label learned from ldp for PE2s next-hop. P routers will know the label bindings for the next-hops as they are running ospf and ldp. As you can see no BGP is running on the P router i.e. P routers only need to how to get to the next-hop.

gorebrush

As Ed says above. Can't really add anything to that!

dppagc

Ok. One question. For the P routers, where does the ip address of the customers go to?
Each ip address is tagged to a label right?
Or am I missng something here?

fredrikjj

dppagc wrote: »

Ok. One question. For the P routers, where does the ip address of the customers go to?
Each ip address is tagged to a label right?
Or am I missng something here?

P routers don't need to carry the customer routes because they (label) switch packets based on the PE router's IP next hops (which P routers do have). Ingress PE looks at a BGP route received from another PE and finds a label and an IP next hop. It adds the vpn label found in BGP, looks at the IP next hop and then adds a second label that will get the packet across the P routers to the other side. The egress PE receives the packet with the VPN label and knows where it's supposed to go because it was the router that advertised that label in the BGP update to the ingress PE.

MPLS Fundamentals is very readable if you want to read a book on it.

dppagc

EdTheLad wrote: »

PE1 will learn the next-hop of PE2 via IGP, when packets are sent to PE2 they will be impose the label learned from ldp for PE2s next-hop.

That is my question. From PE1 to PE2 there are P routers right?
The next hop is learnt via IGP, but wouldnt the labels add on to the load??

networker050184

Depends what you mean by "the load". Yes the Ingres PE will need to push a label, but these headers are easily assembled and pushed. The P routers simply swap labels without needin further route lookup. Without the label a route lookup would be performed independently at each hop.

fredrikjj

If you want to understand why it's not a problem that the packet has additional labels, this is a good article to read: Packet Classification: A Faster, More Generic Alternative to Demultiplexing - The Internet Protocol Journal, Volume 15, No. 4 - Cisco Systems

Essentially, devices look at the entire packet at once and sees the headers as a series of bits, and don't process headers individually. Extra labels are therefore not a problem from a forwarding performance perspective, assuming that the device is designed with that in mind.