Building autonomous network operation using deep learning and AI

Description

The growing adoption of mobility, IoT, and the cloud significantly increases the impact of network infrastructure on enterprise businesses and creates new challenges for traditional human-driven network operation. You’ll learn empirical experiences of using ML, deep learning (DL), and AI to build the first autonomous enterprise network operation solution that provides visibility, troubleshooting, reporting, and maintenance of an enterprise network.

Jisheng Wang explores detailed architecture with an emphasis on measurement, detection, orchestration, and action. The process is to collect a variety of data (e.g., logs, stats, and events) to measure the performance of both infrastructure (e.g., wireless, wired, and internet) and end-user experiences. Based on these metrics, the normal baseline of end-to-end network components, detected anomalies, and real-time incidents is profiled. Spatial and temporal analysis is applied to determine the scope of the impact (e.g., user, site, or organization) and the root cause of the incident. The intelligent system then suggests or takes actions automatically.

He also describes the two ML models built for automated actions. The natural language processing (NLP)-based machine log analysis solution automatically identifies the root cause of machine hardware or software defects using syntactic analysis (term frequency-inverse document frequency (TFIDF) and Doc2vec) and topic modeling. The other is a reinforcement learning-based WiFi radio management solution that automatically tunes radio configuration based on environmental and user dynamics, as well as a user experience-based reward mechanism.

Jisheng shares his experiences and the lessons he learned, including how to use visualization, model interpretation, and feedback to keep humans in the loop while developing machine intelligence; how to build human trust in the step-by-step process of automation, augmentation, and autonomy; and how to accelerate knowledge learning and sharing across enterprises in a SaaS environment without compromising individuals’ privacy information.