A Combinatorial Optimization Framework for the Design of resilient iBGP Overlays

摘要

The Internet is an aggregation of Autonomous Systems (ASes) which exchange network prefixes reachability advertisements using the Border Gateway Protocol (BGP). ASes set up external BGP (eBGP) sessions between the AS border routers (ASBR) of neighboring ASes, while internal BGP speakers establish internal Border Gateway Protocol (iBGP) sessions to learn reachability for external prefixes. In order to avoid loops in the control and forwarding plane, and to ensure complete visibility and path diversity, routers within the same AS must deploy full-mesh BGP sessions, which causes scalability problems, both in the number of sessions and the resources (memory, CPU) consumed by BGP routers. Route Reflection is a widely accepted alternative to improve scalability, but requires careful design, as new issues may be introduced, such as: increased probability of loops, divergence and routing sub-optimality. In our previous work we presented Optimal Route Reflector Topology Design (ORRTD), a combinatorial optimization approach to tackle the problem of designing a consistent and yet optimal iBGP overlay, which minimizes the number of Route Reflectors (RRs), guaranteeing that no sub-optimal route is chosen, i.e., the routes selected with the designated RRs are those that would have been selected if instead of having RRs, the iBGP speakers were fully meshed. In this paper we propose a modification to ORRTD that addresses resilience, i.e., survivability to node or link failures.