BGP-ESOM: BGP Egress Selection Optimization Model Based on Traffic Demand

Abstract

Presently, the most common egress selection algorithm that we used in transit AS (Autonomous System) is hot-potato algorithm. However, the hot- potato algorithm has many problems which have been discussed in some papers. Prior research has considered a range of possible techniques, but has mostly focused on how to make the best static assignment. There is a need for good model and framework that consider how to make the best static assignment and how to adapt to changes (such as intra-domain topology change and traffic change) at the same time. In this paper, we propose a BGP egress selection optimization model (BGP-ESOM), which consists of two phases: the static phase and the dynamic phase. An implementation framework (BGP-RCS) is also developed for it. In the static phase, the assignment is realized by converting the BGP egress problem into the problem of making the best configuration of the IGP (Interior Gateway Protocol) metrics. In the dynamic phase, some algorithms are used to adapt to the topology changes. The main idea is that we use the static phase to make the long-term static optimization, and use the dynamic phase to make the short-term but timely dynamic optimization.