Data Science Techniques for Beginner Network Administrators

Abstract

There is an old saying: "It is easier to wake someone who is sound asleep than to wake who is pretending to sleep." As network administrators, competing with Al systems in the future is not an impossibility. Analyzing events with simple examples like comparing bandwidth, throughput, jitter, and optimizing server resource usage can be exhausting when applied to dozens or hundreds of devices.

For instance, how do we classify when the bandwidth is low in one place but normal elsewhere based on several sample events? Or how do we classify which servers need maintenance and which do not, based on existing data parameters, referring to unsupervised and supervised learning? Classification models can indeed simplify and optimize our work, saving time compared to performing network maintenance one by one, using only Google Colab.

In this context, we have tried using a small of unsupervised and supervised learning to process data from simple network case studies. These data are analyzed for classification, regression, and prediction using Python scripts implemented in Google Colab. This book serves as a starting point to be developed into a continuous series, extending to supervised learning, deep learning, and reinforcement learning, which are still in the testing and writing phase. We hope this new approach benefits network administrators wherever you are.