Jacob:
An MPLS VPN defines an architecture and a set of routing and switching protocols that allows a private enterprise to extend its network across a public domain. Your headquarters building represents a private network that must exhchange routing information and actual user data with your remote sites. The remote sites are private networks that belong to your domain, too.
MPLS allows you to extend your routing domain and use the service provider's Layer 3 network as if it were your own, so to speak. You can route data from private RFC 1918 addresses at each end over a public domain shared by many of the service provider's clients. Furthermore, you can have direct remote site-to-remote site (spoke-to-spoke, so to speak) communication and not have to go through the hub. This is commonly referred to in marketing circles as any-to-any connectivity.
To you, it seems like it's private. None other of the service provider's clients see your routing information nor do they have any knowledge of your topology. So, a VPN is set up per client, and a client can use more than one VPN. Moreover, routes learned from one VPN can be shared by another to facilitate inter-VPN traffic..
You can peer with the service provider using BGP at the CE (customer edge router -- yours) router at the headquarters and the remote sites. And you can then redistribute those BGP-learned prefixes into your IGP, like EIGRP.
You can also run EIGRP, OSPF or RIPv2 with the service provider. The most common application is BGP, though. OSPF presents many complications with routing metrics.
HTH
Victor
