Randolph Yao

TL;DR: It is argued that a key feature of gray failure is differential observability: that the system's failure detectors may not notice problems even when applications are afflicted by them, and should focus on bridging the gap between different components' perceptions of what constitutes failure.

TL;DR: A cost-sensitive ranking-based machine learning model that can learn the characteristics of faulty disks in the past and rank the disks based on their error-proneness in the near future is developed and successfully applied to improve service availability of Microsoft Azure.

TL;DR: A failure prediction technique, which can predict the failure-proneness of a node in a cloud service system based on historical data, before node failure actually happens, is proposed and successfully applied in real industrial practice.

TL;DR: This work formulates the practical scenario as the predictive VM provisioning (PreVMP) problem, and proposes Uncertainty-Aware Heuristic Search (UAHS) for PreVMP, which first models the prediction uncertainty, and then utilizes the predicted uncertainty in optimization.