While it is true that we are in the middle of one of the Artificial Intelligence hypes, it is also true that the combination of unprecedented computation power and data availability, with new variations of well-seasoned Machine Learning algorithms, is dramatically changing the optimization strategies for large ICT industries. Especially, the telecommunications industry has always had to deal with complex systems, stochastic contexts, combinatorial problems, and hard-to-predict users; Machine Learning-aided optimization was just waiting there to be used by telecoms. In this paper, we introduce some basic Machine Learning concepts and discuss how they can be used in the context of telecommunications networks, particularly in wired and wireless networks.
In Automated networking, the use of ML technology can help automate routine network operations involving visualization, analysis, and troubleshooting. It has been a constant industry effort to replace manual network management processes with various ways of automation and now through ML-based algorithms which are indeed the de facto future.
The ML technology relies on various data sources like packet-level telemetry, account log messages, KPIs, and faults data generated by network devices and then analyzed by an ML algorithm. The algorithm then learns how to recognize normal versus abnormal behavior patterns over time, predicting when the element/node might be having issues before they occur.
This whitepaper focuses on various aspects of network automation and how networks can adjust and optimize themselves based on real-time traffic flows, congestions, port failures, high rates of packet drops, configuration changes, software updates, and more without human intervention.