Using Communities to influence default route over MPLS

melcara · ‎06-19-2009

We have a cusomter, who have two internet connections (not part of the BGP config) and two core locations, and about 15 branch offices connected to the cores running MPLS/BGP. What I would like to do, is have half of the branches on ISP1 at Core 1 and the other half on ISP2 at branch 2.

I tried to apply bgp community (additive) so both routes might show in the branches BGP table. No Luck? Any ideas? Also, another option would to tell MPLS, to send all routes to branch, currently only one route is sent to the Branch.

pradeepde · ‎06-25-2009

To set the Border Gateway Protocol (BGP) COMMUNITIES attribute, use the set community command in route-map configuration mode. Use the route-map global configuration command, and the match and set route-map configuration commands, to define the conditions for redistributing routes from one routing protocol into another. Each route-map command has a list of match and set commands associated with it. The match commands specify the match criteria-the conditions under which redistribution is allowed for the current route-map command. The set commands specify the set actions-the particular redistribution actions to perform if the criteria enforced by the match commands are met. The no route-map command deletes the route map.

The set route-map configuration commands specify the redistribution set actions to be performed when all of a route map's match criteria are met. When all match criteria are met, all set actions are performed.

dbass · ‎02-03-2010

Keep in mind that a community string is just a tag used to identify a route that you will take a particular action on. IE: you set a tag on the default at each "core" site, and then at a remote site you identify the route you want to prefer based on that community and set local pref or weight so that it is prefered.

That being said, you have limited ability to manipulate routing inside the providers cloud once the default route is advertised in to it. The cloud will use and advertise to the remote site the default based on BGP best path within the provider network. Also, you cannot send 2 default routes across 1 link and peering session to a remote...only the best one will be used. What you want to do is possible though and you have 3 options:

1. Work with the provider so that they will prefer the default at each POP in the manner that you want to distribute the traffic. So, if you have a remote site in NY and one in ATL and a core in DC and one in MIA. You want the NY site to take the Internet that is in DC, so the provider will configure their network so that the POP that the NY site is connected to will prefer the default route coming out of DC...the opposite occurs for ATL and MIA. Now, not all providers will do this and if they do you are counting on them to do it right, which in my opinion is a little too much reliance on the provider.

2. You peer with your own devices and you will have all the control you will ever want plus support for multicast. You will either need a L2 MPLS network or some type of shared media offering like Metro Ethernet, or you use a GRE hub and spoke design between your routers. This solution has it's own obvious problems, so it's not really an easy choice.

3. You have 2 MPLS clouds. You will have the same control, maybe a little more redundancy (debatable), and a little less complexity and potential problems as you have with option 2. It will also cost you more money as well as longer to implement.

Hope that helps