Horizontal Scaling in Kubernetes Dataset Using Artificial Neural Networks for Load Forecasting
Description
This dataset provides comprehensive data collected from experiments to enhance horizontal scaling in Kubernetes clusters using Artificial Neural Networks (ANNs) for load forecasting. The dataset includes measurements of CPU load, packet reception rates, and the number of pods gathered under various traffic scenarios. These metrics were meticulously collected and preprocessed to train and evaluate the performance of the ANN-based scaling solution compared to the traditional Horizontal Pod Autoscaler (HPA). The dataset is divided into two main parts: training data and test data, each serving a specific purpose in developing and validating the scaling models. The training dataset (TrainData.csv) consists of averaged metrics from multiple experiments, capturing the mean CPU load, mean packet reception rates, and mean number of pods under different stress conditions. This data was collected using a controlled setup with varying numbers of virtual users and replicas, ensuring a robust and diverse set of inputs for training the ANN model. By analyzing these averaged values, the ANN can learn to predict the optimal number of replicas needed to handle specific loads, aiming to improve resource utilization and maintain service quality. The test dataset is split into two components: data collected from the Kubernetes cluster (TestK8sData.csv) and data from the JMeter tool (TestJMeterData.csv). The TestK8sData.csv file includes timestamped records of CPU usage, packet reception rates, number of pods, and additional parameters such as experiment type (HPA or ANN) and CPU thresholds. The TestJMeterData.csv file provides detailed logs of HTTP requests, including response times, success rates, and other performance metrics. Together, these datasets enable a comprehensive evaluation of the ANN-based scaling approach, allowing for comparisons against the traditional HPA method regarding efficiency, responsiveness, and adherence to service level agreements (SLAs).