Contractual relationships between Autonomous Systems (AS) affect inter-domain packet routing and shape the evolution and properties of the global AS-level topology of the Internet. In this talk, I will first describe the problem of inferring AS relationships and then will introduce novel inference heuristics finding customer-to-provider, peer-to-peer, and sibling-to-sibling relationships. I will outline validation results based on a survey with network operators showing inference accuracy between 82.8% and 96.5%. Finally, I will discuss an AS relationships repository we have opened to make our results useful for the community where we archive periodically the Internet AS-level topology annotated with inferred AS relationships.

In the second part of the talk, I will switch to discussing the problem of computing network traffic heavy hitters using limited memory resources. I will briefly introduce the IBM Aurora system, which provides the context of our interest and then I will present an algorithm called Probabilistic Lossy Counting (PLC) for finding network traffic heavy hitters. PLC enhances the well-known lossy counting algorithm using on a tighter error bound on the estimated sizes of traffic flows providing probabilistic rather than deterministic guarantees on its accuracy. Performance comparison experiments show that PLC has between 34.4% and 74% lower memory consumption and between 37.9% and 40.5% fewer false positives than other state-of-the-art algorithms.