Diff: BGPvsOSPF

--- Version 4 

+++ Version 5 

@@ -17,6 +17,13 @@ 

 ||Better administrative controls|| || 

 ||Excellent support for FlapDampening|| || 

 ||Better support for the administrative structure of the network|| || 

- 

+||Allows regional clusters to choose their own routing policy|| 

 

 Now an initial survey of the above would suggest that BGP is the logical choice for the network.. but the con against it is MASSIVE particularly when you consider that we already have these multipath questions being asked see RGsouthern 

+ 

+27/07/05 [DanFlett]: By "less information about specific links within RGs" are we talking about redundant links?  BGP isn't as good at making use of redundant links as OSPF as far as I'm aware.  What I would like to find out is whether BGP will allow more than one AS to represent the same address space - for example - 

+Node A is AS 64500, is a gateway to 10.10.255.0/24 

+Node B is AS 64501, is also a gateway to 10.10.255.0/24 

+This is contrary to the way BGP is supposed to be used, but if it is allowable, it could be quite useful to us - it would mean we could use OSPF in the clusters with a very simple config file, and the border routers would run BGP to speak to the backbone and OSPF to speak to the local cluster - again the config files would be simple. 

+Why do this?  To avoid having to configure iBGP.  BGP gets complicated when you starting having use iBGP to specify the other eBGP speakers in your own AS.  A less complicated setup would be to simply give the eBGP speakers on the borders of the cluster different AS numbers. 

+I'm not sure what sort of effect this config will have - will it cause routing loops within a cluster and will BGP get confused if it sees two ASes representing the same address space?  It would be good to get answers to this question...