Statistical Models for Predicting Scalability and Elasticity in Cloud Applications

Abstract

With the pervasive adoption of cloud applications, the importance of scalability and elasticity has become paramount. This paper delved into the statistical models used for predicting these crucial aspects, evaluating their performance, and subsequently, recommending practices for their implementation. Through an exploration of linear regression models, time series forecasting, Bayesian inference, and ensemble approaches, the research offered insights into their applicability in real-world scenarios, such as e-commerce platforms and streaming services. By comparing model performances using metrics like MAE, RMSE, and R², it showcased the implications of predictive inaccuracies. The study concluded with robust recommendations for the training, deployment, and continuous refinement of these models, emphasizing the balance between predictive accuracy and computational costs.