Load Prediction Techniques in Cloud Environment

Abstract

Businesses and websites have rapidly increased their energy consumption, necessitating the development of data centres tailored to the cloud. Predicting when a system's resources will be needed means you can allocate them more efficiently and save money in the cloud. Predictive accuracy may be increased by classifying loads first. In this research, we offer a new method for predicting future demand for cloud-centric data centres. The Phase Space Reconstruction (PSR) and Extended Approximation-Group Method of Data Handling (EA-GMDH) methods are compared to the Bayesian model for predicting the mean load over a long-term time period. Multi-step ahead CPU load prediction using Support Vector Regression is very stable, i.e., its prediction error increases quite slowly as the predicted steps increase; this is in contrast to a neural network, which predicts the future load based on the past historical data and is distinguished by the presence of hidden layers