The interaction of IGP weight optimization with BGP

Abstract

Link weight optimization is shown to be a key issue in engineering of IGPs using shortest path first routing. The IGP weight optimization problem seeks a weight array resulting an optimal load distribution in the network based on the topology information and a traffic demand matrix. Several solution methods for various kinds of this problem have been proposed in the literature. However, the interaction of IGP with BGP is generally neglected in these studies. In reality, the optimized weights may not perform as well as expected, since updated link weights can cause shifts in the traffic demand matrix by hot-potato routing in the decision process of BGP. Hot-potato routing occurs when BGP decides the egress router for a destination prefix according to the IGP lengths. This paper mainly investigates the possible degradation of an IGP weight optimization tool due to hot-potato routing under a worst-case example and some experiments which are carried out by using an open source traffic engineering toolbox. Furthermore, it proposes an approach based on robust optimization to overcome the negative effect of hot-potato routing and analyzes its performance