Work

Predictive models for adapting microservice-based applications: a comparative analysis

Year: 2025
Type: software
Abstract: Predictive models for adapting microservice-based applications: a comparative analysis Project Description This repository contains the implementation and experimental results of a comparative analysis of statistical, machine learning (ML), and deep learning (DL) algorithms for forecasting microservice time series. The study evaluates ten algorithms: one statistical algorithm, Autoregressive Integrated Moving Average (ARIMA); five deep learning algorithms, Dual-Stage Attention-Based Recurrent Neural Network (DA-RNN), Deep State Space Model (DeepState), DeepAR, Long Short-Term Memory (LSTM), and Temporal Fusion Transformer (TFT); and four traditional machine learning algorithms, Multilayer Perceptron (MLP), Support Vector Regression (SVR), Random Forest (RF), and Extreme Gradient Boosting (XGBoost). For further information about the project, including detailed descriptions of the repository structure, experimental procedures, model hyperparameters, and execution instructions, please refer to the README file available in the project repository. Dataset Description The experiments use 40 time series extracted from production microservices in the large-scale Alibaba cluster deployment available in the Alibaba Cluster Trace dataset. The repository provides the selected time series and supporting information in the following directories: time_series/alibaba/: Contains the 40 time series used in the experiments. other_results/data-descriptions/: Contains descriptive information about the selected time series, including their characteristics, microservice and container IDs, and seasonality. other_results/dtw/: Contains the Dynamic Time Warping (DTW) distances between the selected time series. Installation Create and activate a virtual environment: python3 -m venv venv source venv/bin/activate Install the project dependencies: python -m pip install --upgrade pip python -m pip install -r requirements.txt Project Files Summary of the main repository files and directories: Repository Path Content Description other_results/data-descriptions/ Descriptive information about the selected time series, including their characteristics, microservice and container IDs, and seasonality. other_results/dtw/ Dynamic Time Warping (DTW) distances between the selected time series. other_results/friedman-test/ Results of the Friedman and Nemenyi tests for each microservice metric, used to compare the forecasting performance of the evaluated algorithms. other_results/hyperparameters/ Selected hyperparameter configurations for each forecasting model. other_results/plots/ Plots illustrating the forecasting results of the models across different lag configurations. results/ Mean Squared Error (MSE) and Model Efficiency Time (MET) results for the evaluated models. pickle/ Trained models serialized and stored as pickle files. Generating Experimental Results The following scripts generate the experimental results using the trained models stored in the pickle directory: generate_met_results.py: Generates MET results from models. generate_mse_results.py: Generates MSE results from models. Experimental Environment Hardware Configuration. The experiments were conducted on a Dell Vostro 5490 laptop equipped with an Intel Core i7-10510U processor (4 cores and 8 threads), 16 GB RAM, an NVIDIA GeForce MX230 GPU (2 GB), and Intel UHD Graphics. The system ran Linux Mint 20.3 (Una), based on Ubuntu 20.04 LTS, with Linux kernel 5.4.0-216-generic. Software Configuration. The experimental environment used Python 3.8.10 and the following main software libraries and frameworks: TensorFlow 2.10.0, Keras 2.10.0, PyTorch 1.12.1+cu102, scikit-learn 1.1.2, NumPy 1.23.1, GluonTS 0.10.2, XGBoost 2.1.3, and pmdarima 1.8.5. The complete list of dependencies and their respective versions is available in the requirements.txt file. GPU Acceleration. The LSTM and DA-RNN implementations were executed on the NVIDIA GeForce MX230 GPU using CUDA. All other algorithms were executed on the CPU because GPU acceleration was not supported at the time, either due to limitations of the respective software libraries or hardware compatibility issues. Citation If you use this repository in your research, please cite the associated publication once it becomes available. The bibliographic reference and DOI will be added to this section upon publication. License This project is licensed under the MIT License. See the LICENSE file for details.
Author: unknown, unknown
Language: English
Open Access status: green
OpenAlex ID: W6930671284
Cites:
Cited by:
Related to: 10