PD-GABP — A novel prediction model applying for elastic applications in distributed environment

Abstract

In comparison with other scaling techniques, forecast of workload and resource consumption brings a great advantage to SaaS operations in cloud environment because system knows early and precisely the number of resources must be increased or decreased. However, the prediction accuracy still needs to be improved further even though there are many research works that have dealt with the problem. In this paper, we present a novel prediction model, which combines periodicity detection technique and neural network trained by genetic-back propagation algorithm to forecast the future values of time series data. The model is experimented with real workload dataset of a web application. The tests proved significant effectiveness of the model in improving the prediction accuracy. Our model thus can enhance the performance of applications running on cloud and distributed environment.