Designing An Optimal And Resilient Ibgp Overlay With Extended Orrtd

The Internet results from interconnecting several thousands of Autonomous Systems (ASes), which are networks under a single administrative domain such as: corporations, service providers, universities, and content providers, among others. To ensure global communication between end users, it is necessary that routers of every AS get to learn all IP addresses in this immense and extremely decentralized network. The Border Gateway Protocol (BGP) is the responsible of learning and distributing that reachability information among ASes in the form of groups of addresses (a.k.a. prefixes). Unlike other routing protocols, BGP routers communicate through administratively set point-to-point BGP sessions over TCP. BGP sessions are either external (eBGP, between routers of different ASes, a.k.a. Border Routers or ASBRs) or internal (iBGP, between routers whit to the same AS). While eBGP is needed to exchange reachability information among ASes, iBGP makes it possible for internal routers to learn prefixes necessary to forward IP packets to the appropriate ASBRs. To make sure that the whole information is learnt and no traffic deflection occur, a full-mesh of iBGP sessions among routers within each AS can be used, which causes scalability issues. Route Reflectors (RR) is a mechanism to improve performance, but designing a: correct, reliable and consistent iBGP overlay of sessions whith RRs is a delicate, far from easy task for ASes engineers, even though several popular heuristics are common practice. In previous works we proposed combinatorial optimization models to design consistent and resilient BGP overlays, when only non-Border-Routers are eligible for RRs. The present work extends previous models to allow any router (including Border Routers) to be Route Reflectors, which matches better to some application contexts.