cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
1007
Views
5
Helpful
5
Replies

IGP for Loopbacks Only Vs IGP for All the Interior routes ?

Hi !

can anyone help me with this one ?

there are networks which use igp for all of the igp routes , connected and static, and they use bgp only for exterior routes [ like from other ebgp sessions ].

and there are networks which use the igp for advertising only the loopbacks , and over those loopbacks run the ibgp sessions which contains all the igp and egp routes.

i wanted to know if and what could be the cause to prefer building a network like each ? is there a real reason ?

Thanks

2 Accepted Solutions

Accepted Solutions

Marwan ALshawi
VIP Alumni
VIP Alumni

very good question

for the case one where you have IGP internally only and BGP [eBGP] with other routing domains

this approach is a good option where you have one BGP exit point and you do not need your internal network to know all the eBGP routes such as Internet routing table ovr 300K routes in this case you injuct default route from the BGP routers to the internal IGP and the IGP used to maintain internal reachability between networks

for the other case where you have iBGP internally and eBGP with outside domains

this is useful for redundant EBGP routers, and also for some BGP big networks where you have you BGP AS as transit AS ( other BGP trsit you network to go to another BGP domain ) same concept to service providers as well

this is from deisgn point of view

from technical point of view any approach of the above can have reahcability and routing exchange but gain depends on the policies, size of the routing table and network you can decide always keep it simple better

HTH

if helpful Rate

View solution in original post

Jon Marshall
Hall of Fame
Hall of Fame

Just to add to Marwan's very good post (+5 marwan), since the advent of MPLS a lot of this has changed in the sense that EBGP/IBGP is only run on the edge of the network and internally only an IGP to provide reachability between the PE routers which run BGP.

If you do run IBGP internally as an IGP then a requirement is to have a full mesh due to IBGP route propogation rules. Using route-reflectors and confederations are ways of mitigating have to run a full mesh.

Jon

View solution in original post

5 Replies 5

Marwan ALshawi
VIP Alumni
VIP Alumni

very good question

for the case one where you have IGP internally only and BGP [eBGP] with other routing domains

this approach is a good option where you have one BGP exit point and you do not need your internal network to know all the eBGP routes such as Internet routing table ovr 300K routes in this case you injuct default route from the BGP routers to the internal IGP and the IGP used to maintain internal reachability between networks

for the other case where you have iBGP internally and eBGP with outside domains

this is useful for redundant EBGP routers, and also for some BGP big networks where you have you BGP AS as transit AS ( other BGP trsit you network to go to another BGP domain ) same concept to service providers as well

this is from deisgn point of view

from technical point of view any approach of the above can have reahcability and routing exchange but gain depends on the policies, size of the routing table and network you can decide always keep it simple better

HTH

if helpful Rate

Thanks jon and marwan !

so if i inject only default route to the network , it would cause traffic to go to the bgp router , and there to do the calc to see if it has a better path to go to . and if i have ibgp then it would choose the supposdly best path in the interior network as well , even before it would go to the border router , correct ?

because if i have two bgp border routers , the igp will not be aware of the best path to the prefix , but only that of the preferred default route ...

thanks guys !

Not sure i entirely follow.

If you have network with a EBGP router on the edge and then you run an IGP inside then yes if you inject a default-route into the IGP from the BGP router all traffic going out of that network woud go to the BGP router, providing there were no other exit points advertising better routes.

If you had a network with IBGP as your interior routing protoco but still only exit point then the traffic woud still end up going to the EBGP router on the edge.

Whether you have an IGP such as OSPF/EIGRP or your run IBGP the answer to you question is more to do with whether you have multiple entry and exit points in your network as Marwan suggested. So if you ony have one exit point within the network then it doesn't make much difference how you do it.

If you have multiple exit points in your network and you want to influence the exit path that is taken this is where IBGP comes in but you can still run an IGP for reachabiity information with the network.

Jon

Hi

Just to add to Jon's post, the best path selection always determied by the used routing protocol either BGP or IGP such as OSPF

but each routing calculate the best path diffrently fo example ospf uses costs of the links on th epath while BGP uses many attributes to consider best path

again as mentioned earlier in this discussion when you have a network to be designed for routing you need to know the over all requriments and topology before you make the call if you are going to use BGP or IGP with BGP for example

HTH

Jon Marshall
Hall of Fame
Hall of Fame

Just to add to Marwan's very good post (+5 marwan), since the advent of MPLS a lot of this has changed in the sense that EBGP/IBGP is only run on the edge of the network and internally only an IGP to provide reachability between the PE routers which run BGP.

If you do run IBGP internally as an IGP then a requirement is to have a full mesh due to IBGP route propogation rules. Using route-reflectors and confederations are ways of mitigating have to run a full mesh.

Jon

Review Cisco Networking products for a $25 gift card