Fault Prediction in Server-Side of Web-Applications by Using a Software System and Long Short-Term Memory Neural Networks

This article postulates architectural design and development of a failure prediction system for server-side web applications based on Long Short-Term Memory (LSTM) recurrent neural network. Importance is given to model building and analyzing important aspects of neural network design. Additionally, the paper stresses the design of a system that collects application performance data. The architecture of the system, the way of storing the data and the choice of metrics to be used have been chosen. The paper introduces a system predicting failures in web server applications based on a Long Short-Term Memory (LSTM) recurrent neural network. The research aims at creating a model capable of analyzing server performance metrics (e.g., CPU utilization. throughput, response time) and predicting possible failures based on collected time series. The author describes the necessary steps to build such a system, including the selection of a database management system, the definition of metrics, and presents the code that forms the basis for performing the prediction using the Python programming language. The paper concludes with the main results and conclusions of the research.