Probabilistic and reactive fault diagnosis for dynamic overlay networks

摘要

Overlay networks have emerged as a powerful and flexible platform for developing new disruptive network applications. The attractive characteristics of overlay networks such as routing flexibility and overlay topology dynamics bring to overlay fault diagnosis new challenges, which include the dynamical overlay symptom-fault correlation, multi-layer (i.e., underlay vs. overlay) abstraction, and unregulated overlay symptoms. To address these challenges, we propose a novel user-level probabilistic and reactive fault diagnosis technique, called ProFis for overlay networks, which can seamlessly integrate passive and active fault reasoning to develop an optimal fault diagnosis framework. ProFis uses observable overlay symptoms as reported by overlay applications to dynamically correlate overlay symptoms and faults. ProFis diagnoses overlay faults passively and selects optimal actions (i.e., with the least cost) to enhance the passive diagnosis whenever necessary. Our evaluation study shows that ProFis can efficiently (i.e., low latency) and accurately localize the root causes of overlay faults, even when symptom loss rate is high.