Issue link: https://www.e-digitaleditions.com/i/1414667
26 I ICT TODAY Many DCI providers reserve one or more fiber strands in their cables, specifically for the purpose of monitoring cable integrity. By taking advantage of virtualized or con- tainerized QoS and QoE monitoring solutions, the DCI provider can generate their own QoS and QoE metrics to correlate with their fiber monitoring metrics. Traditional round-robin fiber monitoring across large fiber counts means an individual fiber may only be mon- itored as little as once every 1.5 hours—sufficient to detect longer term loss degradation, but not necessarily instan- taneous faults. However, monitoring QoE and QoS KPIs and metrics continuously provides a real-time alerting capability, after which the DCI operator could preempt the round-robin testing to perform fiber level troubleshooting on the specific fiber to determine if it is at fault. This is not limited only to dark fiber testing. By uti- lizing a coupler to insert a test wavelength into a specific lit fiber, this same multi-layer assurance can be done on a customer service level and a fiber cable level. to monitoring and assuring services would rely on optical layer monitoring, as well as QoS (Layer 2, 3) monitoring, combined with ML, to automate the detection and correlation of faults within the DCI provider's domain. THE BUSINESS OPPORTUNITY OF MULTI-LAYER AI DRIVEN ASSURANCE An AI-driven multi-layered approach to fiber interconnect service assurance helps providers deliver a more resilient DCI service, avoid punitive impacts of SLA breaches, while also creating additional business opportunities and competitive advantages, including: Integrity—Break and Breach Detection Continuous, cloud orchestrated OTDR monitoring of fiber interconnect services drastically reduces the time taken to manually test individual dark or lit fibers, allowing operators to monitor many more fibers than previously possible. AI AND ML FOR ACTIONABLE INSIGHTS The risk of a multi-layer monitoring solution is the amount of performance data generated and the need to identify anomalies in this data—and then correlate these anomalies to identify the actual root cause. The overwhelming volume of data can quickly lead to alarm fatigue. This, in turn, can lead to operations paralysis, where it becomes almost impossible to efficiently deter- mine the correct priority of actions needed to maximize repair and reconfiguration activities to address the broadest number of impacted customers, while minimizing financial risk due to SLA penalties and loss of brand reputation. Luckily, this is precisely where artificial intelligence (AI) and machine learning (ML) shine. Done correctly, ML algorithms can be trained to identify anomalies in the live KPI data streams and correlate them against other detected anomalies from a wide variety of seemly disparate sources. These include the optical layer, QoS and QoE KPIs, as well as sources like planned mainte- nance activities, operations support system (OSS) or business support system (BSS) data, weather or seismic events, and even large scale social or sporting events which may drive abnormal traffic patterns. For DCI, a multi-layered AI or ML-driven approach to monitoring and assuring services would rely on optical layer monitoring, as well as QoS (Layer 2, 3) monitoring, combined with ML to automate the detection and correlation of faults within the DCI provider's domain. Artificial intelligence could then be leveraged to provide automated or orchestrated changes to the network and/or service topology to alleviate end cust- omer impacting situations, or on a less ambitious basis, to provide guidance and advice with regard to potential solutions and optimal courses of action. For DCI, a multi-layered AI or ML-driven approach