Sunday, December 21, 2008

MPLS Deployment reasons

1) Faster convergence, in the old days that was a valid reason due to the relatively complex forwarding task that required more resources then Label forwarding. Today non relevant

2) RFC 1483 the newer 2684 AAL5 ATM Adaptation Layer 5 the implementation of IP over ATM

3) BGP Core Free on the SP network, as with MPLS lookup is done based on Labels and not destination address there is no need to have the BGP table in the Core for external prefixes lookup's. this is a massive change from the requirement that Every router in the core must have BGP enabled (cpu and memory intensive load) to only the edge router have BGP enabled mean higher performance and capability.

Note: edge routers still need to have the BGP routing tables, edge routers are translating between ip routing decision to label based decisions.

4) Deployment scalability, when we face with large scale client to deploy (client that connect with 50 - 300 sites and more) we need to have under the consideration the deployment scalability and management, with that in mind 2 models of deployment are optional:

a) VPN Overlay - creating a point to point connection over the SP network, can be achieved in layer 1, 2 or 3.

Layer 1 TDM E1 T1... Layer 2 ATM FR... Layer 3 GRE, IPIP...

b) Peer to Peer - creating a connection between sites trough the ISP and with him, what I mean is that the SP need to join the client network and to achieve client privacy the SP need to manage acl's and routing updates, not very scalable and a lot of overhead. notice that in addition to the disadvantage for the SP on the additional management overhead and complexity there is the client control (doesn't have any) of his layer 3 network trough the SP.

With MPLS the VPN allow the Peer to Peer bad model to have advantage over the Overlay model where in the MPLS we use VRF Virtual Routing Forwarding separators between each network and the configuration is done only on each new site. meaning that if I am an SP and I have 3 client (Cisco, Microsoft, Verizon) each vrf will have a unique color vrf Cisco, vrf Microsoft and vrf Verizon, and to join a new branch is only to color this branch traffic accordingly, so the main work is done in the initial design and implementation and any new addition is actually very simple to add.

5) TE - traffic engineering is a small phrase for a very big spectrum of options, normally traffic routing is decided at each point separately and usually the best route is chosen according to the shortest path to destination, using TE we can make the routing decision based on multiple criteria options. allowing the traffic to fully utilize network capability.

FRR - Fast ReRouting is a very good feature that allow you to detect and reroute based on router availability in less than 50ms Very important in high sensitive traffic like VoIP.

2 comments:

Ahenning said...

Hi Shiran, does this post indicate that you might be pursuing SP? :D

Ahenning

cciep3 said...

Well, Yes!

but much slower then before as I do not have much time