Issue link: https://www.e-digitaleditions.com/i/1414667
October/November/December 2021 I 27 With the ability to test more fibers more frequently, the physical and optical integrity of a much higher proportion of the network can be assured in near real time. Additionally, by using ML techniques, this continuous stream of test data can be automatically categorized and classified. Combined, rapid fiber testing and ML correlation and classification enables providers to quickly detect and identify fiber cuts, as well as unexpected perfor- mance degradations associated with routine main- tenance, cable manipulation often seen in road construction activities, and even potential security breaches such as tapping. Furthermore, providers can continuously monitor for unexpected changes in their dark fiber assets, often associated with illegal use. Faster, More Confident Redundancy When DCI failures do occur, time is of the essence. With near real-time performance data of fiber assets, DCI providers can provide faster failover for their customers' services without the need to conduct a test of the fiber before initiating the switch. By including topological data, the assurance solution can guarantee redundant links are geographically and physically independent over an entire route, avoiding any single point of failure issues. AI-Driven Root Cause Analysis When facing critical outages, providers often rely on teams of highly qualified cross-domain experts, urgently convened to determine the root cause. While effective, these tactical war rooms often take considerable time that incur delays, cost, and reputational impact. The ability to determine patterns rapidly in multi-dimensional data sets is the ideal domain for ML. Machine learning-driven service assurance solutions correlate multi-dimensional data across internal domains and silos, along with additional third-party and external data sources to determine probable root causes in a fraction of the time taken by a war room. The increased speed at which the root cause can be triaged leads to lower overall cost of identification, reduced mean time to resolution (MTTR), and increased customer satisfaction. Fix Not Find Even with the increased speed and coverage an AI-driven assurance solution provides, when a network fault need operations' attention, having an accurate identification of its location is a critical aspect of MTTR reduction. In the absence of this information, additional truck rolls and testing are required to pin-point the location, determine the nature of the fault, and supply the parts. By correlating QoE and QoS KPIs with fiber moni- toring KPIs and topological information, an AI-driven, multi-layered assurance solution can provide the exact nature and location of a fault, thereby eliminating the cost of unnecessary truck rolls. This "fix not find" para- digm eliminates cost associated with in-field diagnostic testing and minimizes the MTTR, ensuring SLA penalties are minimized or avoided altogether.