Analysing global triangle inequality violations due to Internet eXchange Points for future overlay networks

The growth of peering at Internet eXchange Points (IXPs) by Autonomous Systems (ASes) has led to opportunities for the creation of newer architectures for improved inter-domain routing. This peering infrastructure is now a critical component of the Internet ecosystem and may be exploited to identify global routes with lower end to end latencies. In this work we propose a measurement framework to infer triangle inequality violations (TIVs) in the Internet delay space which are created due to worldwide peering. We then study and analyze properties of this artifact of Internet routing policies and motivate the potential for the creation of future overlay networks based on these TIVs. We observe popular peering links provide alternate paths between a set of end hosts which are more efficient and are thus responsible for their creation. By carrying out a graph based study of the underlying topology of TIVs, we provide a theoretical basis of indicating the common subset of good peering links decreasing overall end to end delays. We observe the presence of a significant percentage of alternate paths outperforming the default route through an IXP in all major regions. Even though IXPs are set up primarily with economic benefits in mind, we conclude that there is a huge potential for improvement with more efficient routing and better planning at the primary IXPs around the world.