Harold Ritter Mon, 03/02/2009 - 10:15
User Badges:
  • Cisco Employee,

Kaustubh,


It is usually not recommended to redistributed iBGP routes to IGP as they are usually used to advertise the same routes.


It is required to configure "bgp redistribute-internal" under the BGP process to allow the iBGP routes to be redistributed into the IGP.


Regards

kaustubhchavan Mon, 03/02/2009 - 11:46
User Badges:

If suppose i am redistributing all routes from ISP in to the IGP then have to redistribute my IGP internal routes to isp's router ?


Is tht my statement is correct ?

if correct thn how would it be because it is my private n/W.


Please explain with brief digram


Thanks & regards,

Kaustubh

adamclarkuk_2 Mon, 03/02/2009 - 11:54
User Badges:
  • Silver, 250 points or more

Hi


You definitely do not want to redistribute a complete BGP table into your IGP. BGP is designed to operate at the edge of your network. You want to pass as few prefixs to your ISP as possible (most ISP's wont accpet less than a /21 or a /20), ie if you have a /19 from RIPE,ARIN or APNIC then you do not want to send they all the subnets you are using in your network just the /19. On the other hand, your IGP does not need to know about every prefix on the internet, instead just pass a default route back into your IGP domain so all your routers default to the edge.


To summarize, you need BGP as a demarc between your network and the outside world.

Giuseppe Larosa Mon, 03/02/2009 - 13:07
User Badges:
  • Super Silver, 17500 points or more
  • Hall of Fame,

    Founding Member

Hello Kaustubh,


the approach to redistribute internet routes into an IGP is not feasible nowdays:

the internet routes of a full BGP table are simply too many more then 270,000 to be able to import and handle in whatever IGP you use that has capability to handle 5,000 to 10,000 routes.

What you can do is to inject a default route in your IGP to allow outbound traffic to reach the border router.

This can be done on the different routing protocols and is a good choice with minimal effort and load for the IGP.


if you have multiple exit points = multiple border routers you can run iBGP between them.

Depending on the topology you need to run iBGP over all routers involved in the paths between border routers.

In this case all other routers need just to point to one router in the iBGP mesh to have the traffic routed to the best exit point.


To be noted that MPLS has introduced an additional plane of signalling that allows a router to forward traffic even if it doesn't know how to route it.


iBGP + MPLS provides this great advantage.


So MPLS has greatly relaxed the requirements for iBGP full mesh and in general about the routing information distribution inside a private network.


Hope to help

Giuseppe



Actions

This Discussion