Solved: why use MPLS and benefits

carl_townshend · ‎04-18-2012

Hi all

I would like to know a solid reason for using MPLS for ISP's, I have read some info on this and cant see what the major benefits are, as you have to already have your fundamental network in place, for example a mesh of routers using, say HDLC, frame relay, ethernet. they all can see each other via IP, right. so after this we then enable MPLS switching.

what exact benefits are we getting here ? , am i right in saying the all the devices need to be able to see each other via Layer 3 IP first, then the labels get distributed and assigned to each prefix.

can someone please explain, why SP's are using this ?

cheers

Peter Paluch · ‎04-18-2012

Hello Carl,

Personally, I see here a couple of reasons that can be attributed either to simplifying the network configuration and operation, or to offering services that weren't possible with plain IP:

Simplified BGP configuration and operation. Instead of having each backbone router run BGP and carry both internal and external routing table, only ISP's edge routers (the PE routers) run BGP and are peered together. Internal backbone routers (the P routers) run only an IGP protocol but do not need to run the BGP. This both greatly simplifies the BGP configuration (think peering tens or hundreds of BGP speakers - even using confederations or BGP route reflectors is going to be tedious in large deployments) and also decreases the load on internal P routers.
Possibility to perform traffic engineering. Instead of always relying on the shortest path, you can engineer diverse paths through your network, utilizing even longer-cost routes to the same destination, and map the traffic onto these so-called Traffic Engineering tunnels. You can therefore better utilize the available bandwidth in your network, as you are avoiding the bottlenecks of a single path. The granularity of this mechanism can be down to a hop-by-hop manual definition of the route should you so desire.
Fast Reroute mechanisms. This is really an extension of the traffic engineering. You can pre-define backup paths that provide an alternate routes (detours) around a protected link or node. When a router detects a failure on a link or an outage of a neighbor, it can immediately start using the pre-engineered repair path, without having to wait for routing protocol convergence. This can significantly lower the duration of outages in case of link or node failures.
VPN services. Probably the largest field of applications in MPLS. The features of MPLS allow for scalable creation of diverse VPNs, be it L3 VPNs (the provider, to a certain point, participates in customer's routing), L2 VPNs (the provider simply provisions a pseudowire interconnecting two sites, or provides VPLS - multipoint L2 VPN service), Any Transport over MPLS (a class of L2VPN applications that allows to carry frames through the provider's network without requiring the P routers to actually understand their format). The capabilities of these VPNs also allow for creating multicast-enabled VPNs. A somewhat special class of VPN applications is the Carrier Supporting Carrier, where a larger carrier can offer data transportation services to a smaller regional carrier, with the smaller carrier still being able to provide (almost) all MPLS-based services described above. One of the results of using MPLS VPNs is the possibility to actually separate the address spaces of the customer and the provider, disallowing the customer to communicate with provider's devices on a packet forwarding level (a customer enclosed into a VRF can not talk to anything outside the VRF system assigned to him).

I am sure that other friends here will add another great uses of MPLS I have missed. I have not described technologies like 6PE or 6VPE as they are basically similar to deploying basic MPLS-supported BGP routing and L3 VPNs, but yeah, they're very nice! Also, I personally like the idea of MPLS-TP (Transport Profile) that allows a centralized node to control the operation of all MPLS Label Switching Routers in the network and alleviating the need of running LDP, IGP, RSVP, etc. on these devices altogether - although this field of MPLS applications is still in its infancies to my best knowledge.

Best regards,

Peter

View solution in original post

Peter Paluch · ‎04-18-2012

Hello Carl,

Personally, I see here a couple of reasons that can be attributed either to simplifying the network configuration and operation, or to offering services that weren't possible with plain IP:

Simplified BGP configuration and operation. Instead of having each backbone router run BGP and carry both internal and external routing table, only ISP's edge routers (the PE routers) run BGP and are peered together. Internal backbone routers (the P routers) run only an IGP protocol but do not need to run the BGP. This both greatly simplifies the BGP configuration (think peering tens or hundreds of BGP speakers - even using confederations or BGP route reflectors is going to be tedious in large deployments) and also decreases the load on internal P routers.
Possibility to perform traffic engineering. Instead of always relying on the shortest path, you can engineer diverse paths through your network, utilizing even longer-cost routes to the same destination, and map the traffic onto these so-called Traffic Engineering tunnels. You can therefore better utilize the available bandwidth in your network, as you are avoiding the bottlenecks of a single path. The granularity of this mechanism can be down to a hop-by-hop manual definition of the route should you so desire.
Fast Reroute mechanisms. This is really an extension of the traffic engineering. You can pre-define backup paths that provide an alternate routes (detours) around a protected link or node. When a router detects a failure on a link or an outage of a neighbor, it can immediately start using the pre-engineered repair path, without having to wait for routing protocol convergence. This can significantly lower the duration of outages in case of link or node failures.
VPN services. Probably the largest field of applications in MPLS. The features of MPLS allow for scalable creation of diverse VPNs, be it L3 VPNs (the provider, to a certain point, participates in customer's routing), L2 VPNs (the provider simply provisions a pseudowire interconnecting two sites, or provides VPLS - multipoint L2 VPN service), Any Transport over MPLS (a class of L2VPN applications that allows to carry frames through the provider's network without requiring the P routers to actually understand their format). The capabilities of these VPNs also allow for creating multicast-enabled VPNs. A somewhat special class of VPN applications is the Carrier Supporting Carrier, where a larger carrier can offer data transportation services to a smaller regional carrier, with the smaller carrier still being able to provide (almost) all MPLS-based services described above. One of the results of using MPLS VPNs is the possibility to actually separate the address spaces of the customer and the provider, disallowing the customer to communicate with provider's devices on a packet forwarding level (a customer enclosed into a VRF can not talk to anything outside the VRF system assigned to him).

I am sure that other friends here will add another great uses of MPLS I have missed. I have not described technologies like 6PE or 6VPE as they are basically similar to deploying basic MPLS-supported BGP routing and L3 VPNs, but yeah, they're very nice! Also, I personally like the idea of MPLS-TP (Transport Profile) that allows a centralized node to control the operation of all MPLS Label Switching Routers in the network and alleviating the need of running LDP, IGP, RSVP, etc. on these devices altogether - although this field of MPLS applications is still in its infancies to my best knowledge.

Best regards,

Peter