Frequently Asked Questions

Q—I have a SF hub and a SJ hub. Do you think it is better to separate them into different ASs and run BGP instead of running an IGP in between?

A—This doesn't sound like a candidate for segmentation via BGP. Remember that even though segmentation gives better hierarchy and control, it introduces more routing policies dictated by the BGP behavior. In small networks such as yours, you could achieve the same stability by running an IGP.

Q—I do not have enough BGP peers to justify using route reflectors. What happens if I use them anyway?

A—You will achieve normal routing. You just need to understand that with this model, you rely on centralized routers for running BGP sessions. The RR has to do more processing, and it becomes a single point of failure. Hence, you have to do more provisioning for redundancy. You also will have to deal with other issues such as peer groups and attribute modification, as described in this chapter. If you think that the overhead is not an issue, configuring RRs is no problem.

Q—With confederations, an EBGP external route is more preferred than a confederation external route. Does that mean that I can never use another sub-AS as an exit point?

A—No. You could always use attributes such as local preference to prefer whichever exit point you want.

Q—Because local preference is not passed between ASs, it won't be passed between sub-ASs inside a confederation, correct?

A—That is not true. Using additional configuration, the sub-AS will know that it is talking to an external peer inside a confederation and will maintain all attributes that are normally maintained by IBGP.

Q—I need to configure route reflectors, but the current software on my routers does not support it. Do I need to upgrade all my routers at the same time?

A—No. You only need to upgrade the routers that will become RRs. Other routers will behave as any conventional IBGP speaker. This will help you migrate your network to the new design in a structured way.