Research

Large-scale service environments rely on autonomous systems for remediating hardware failures efficiently. In production, the autonomous system diagnoses hardware failures based on the rules that the subject matter experts put in the system. This process is increasingly complex given new types of failures and the increasing complexity in the hardware and software configurations.

In this paper, we present a machine learning framework that predicts the required remediations for undiagnosed failures, based on the similar repair tickets closed in the past. We explain the methodology in detail for setting up a machine learning model, deploying it in a production environment, and monitoring its performance with the necessary metrics. We also demonstrate the prediction performance on some of the repair actions.