Monitoring and Data Analytics for Optical Networking: Benefits, Architectures, and Use Cases

Machine Learning for Optical Network Security Monitoring: A Practical Perspective

“…Autonomous network operation relies on closing the loop between collecting the telemetry data, analyzing this data by applying different examination and interpretation functions, and performing actions necessary to maintain high network performance. This loop is referred to as the Observe-Decide-Act loop in [10], Observe-Analyze-Act loop in [7], and Collect-Analyze-Test (CAT) loop in [5]. The work in [10] focuses on increasing the efficiency of network resource usage by estimating Quality of Transmission (QoT) and applying ML techniques for margin reduction, along with experimentally demonstrating dynamic connection provisioning and rate adaptation under fiber/amplifier and Reconfigurable Optical Add-Drop Multiplexer (ROADM) aging, as well as frequency correction.…”

“…The work in [10] focuses on increasing the efficiency of network resource usage by estimating Quality of Transmission (QoT) and applying ML techniques for margin reduction, along with experimentally demonstrating dynamic connection provisioning and rate adaptation under fiber/amplifier and Reconfigurable Optical Add-Drop Multiplexer (ROADM) aging, as well as frequency correction. In [7], the authors investigate similar use cases and discuss key requirements, advantages and drawbacks of centralized, distributed and hierarchical monitoring and data analytics capabilities, indicating main issues to be addressed before the potential of these tools can be fully utilized. In [5], key requirements on network diagnosis are examined from an operator's perspective on use cases related to 5G, optical transport disaggregation and multioperator orchestration, detailing on the role of the CAT loop as an enabler of truly autonomous, programmable networks.…”

“…In order to sustain the evolution towards flexible, programmable and autonomous systems, optical networks should support autonomous diagnostics and operation [5]. Integrated telemetry and network analytics [6] are critical for realizing the Observe-Analyze-Act control loop and improve network performance, cost efficiency, and security [7]. The recent proliferation of ML techniques into numerous aspects of optical networking has brought forth new and powerful methods for cognitive and automated management of optical network security.…”

Telemetry in Disaggregated Optical Networks

“…The detailed and combined monitoring of optical connections behavior, along with the physical parameters monitoring disclosed by optical devices (e.g., coherent receivers) are considered essential nowadays to increase network awareness, to identify soft failures, to detect correlation between anomalous events, to localize and to predict/forecast malfunctioning devices before their actual fault [1]. For example, the study of the received spectrum at Optical Spectrum Analyzer (OSA) placed within specific devices and correlations among such spectra may reveal WSS filter malfunctioning (e.g, filter shift, tightening) [2].…”

Machine Learning for Cognitive Optical Network Security Management

“…Network operators still lack the tools to achieve cognitive optical network security management, leaving important aspects of attack prevention, detection and remediation uncovered. Such a framework, supporting autonomous security diagnostics and management, is critical for realizing the Observe-Analyze-Act control loop to improve network performance and cost efficiency [2].…”