AI Summary & Key Takeaways
AI InsightThe model, trained with hyperparameter optimization, achieved an overall accuracy rate of 98.85%; it performed particularly well in the “Normal” and ‘Weak’ classes, but revealed an area for improvement in the “High Load” class with a recall value of 0.88, indicating class imbalance [4].
📌 3 Key Takeaways
Abstract
While microservice architectures offer significant advantages such as flexibility, scalability, and ease of maintenance in software systems, performance analysis becomes a complex process due to their distributed and heterogeneous structures [1, 2]. This study addresses the development and evaluation of a Convolutional Neural Network (CNN)-based model aimed at automatically classifying the performance states (“Normal,” “High Load,” “Poor”) of microservice systems. Performance metrics (CPU usage, memory usage, response time, error rate) were obtained from a simulated dataset and converted into a 2D matrix format suitable for the CNN model's input [3]. The model, trained with hyperparameter optimization, achieved an overall accuracy rate of 98.85%; it performed particularly well in the “Normal” and ‘Weak’ classes, but revealed an area for improvement in the “High Load” class with a recall value of 0.88, indicating class imbalance [4]. The main contribution of this work is adapting CNN's spatial feature extraction capability to microservice analysis by processing performance metrics as an “image.” This innovative approach offers an alternative solution to traditional methods [5]. Future work may explore data augmentation techniques or more advanced network architectures to address class imbalance and test the model's generalization ability on real-world microservice systems [6, 7].