Route reflection topology planning in service provider networks
Viviana Solla, Gabriel Jambrina, Eduardo Grampín · 2017
The Border Gateway Protocol (BGP) has been deployed more than 25 years ago to cope with Internet scalability. BGP is a path-vector, policy-routing protocol which is used to exchange reachability information among Autonomous Systems (ASes), enabling service providers to express their policies, by means of the attachment of several attributes to network prefixes. On any given Autonomous System (AS), external routing information is received at Border Routers (ASBRs), and must be disseminated to internal routers in the AS in order to determine exit gateways for internal (or transit) traffic. It is well-known that a full-mesh of internal BGP (iBGP) sessions is needed to make sure that every BGP speaker has complete visibility of routing alternatives, and therefore, has the ability to take the right choice for any given IP prefix, but unfortunately this solution scales poorly and impose large CPU and memory requirements for BGP routers. Route Reflectors (RR) have been adopted to tackle down the iBGP scalability problem, introducing a new set of problems, including lack of route diversity, non-optimal routing, and route deflection, among others, as extensibly described in the literature. Therefore, the Route Reflection Topology Planning has been a problem of outermost importance, particularly for transit providers. A number of algorithms and heuristics have been proposed, but, confronted with the challenge of designing a particular iBGP network, few tools exist to assist the network operator. In this paper we present RRLOC++, an evolution of our RRLOC tool, which enable the implementation and testing of RR topologies.