BGP enables you to have the best path to networks you are trying to get to, it also allows you to get redundancy if one link were to go down, or one network were not to be reachable via one of your links....
so, if you have router A in one datacenter connected via BGP to one provider
and router B in another datacenter connected via BGP to another provider
how do you utilize the best path...all you have now is a connection with a software program.
also, you must have your own IP space to utilize it as well. your provider may give you some, but then you will have to get your provider's approval to have another network announce those IP routes to the internet...
the two routers, must talk to each other over an internal Routing protocol such as OSPF and then run IBGP between them, so the routers A and B could choose, which route to take.....
so, straight forward, the answer is yes, but it would be a waste of time and money unless you had your two routers actually connected via bgp as well....at least I do not see the benefit in having bgp on two routers not directly connected.
I do not see the benefit in having bgp on two routers not directly connected.
Viper36: The theoretical benefit has to do with ClusterMania's alias
While the servers at the sites would not get the benefit of best path selection if both routers were not directly connected, people trying to send traffic to them would. Whichever data center is closer AS-PATH wise would get the traffic.
The real key here is that I think ClusterMania is trying to get geographic redundancy here If one of the data centers goes off the air, well, it'll stop advertising and everyone on the planet will start sending all of the traffic to the remaining data center. This gives you _really_ cheap -- implementation and software wise -- geographic redundancy.
Of course, this does cause potential issues -- if a routing change causes one AS-PATH to suddenly become longer than another for an ISP, then all of the traffic coming from that ISP will instantly be re-routed to the other data center. All clients affected by the routing change who were connected at the time of the routing change are going to get a lovely 'Connection reset by peer' message every time the routing changes in this manner.
If you can be reasonably sure the routing will be stable, this isn't a problem, but is one of the major reasons most geographic diversity cluster systems play games with DNS instead of routing protocols.
A local loop is just a point-to-point line between two locations. It's going to cost you based upon how much bandwidth you need and how far you need to go. As such, pricing varies wildly depending on your circumstances.
That's something you're going to have to ask whichever telco you plan on using.