Dynamic Cloud Resources Allocation

Abstract

Many software companies have clients that use Microsoft Azure services. Clients may have varying needs for resources, so Microsoft Azure has a very dynamic feature called elastic pool that allows resources to expand and shrink automatically on demand. However, this dynamic feature is very costly both for the clients and the software companies. Thus, there is a growing need to be able to predict the usage ahead of time on daily basis. In this paper we propose and develop an intelligent usage prediction model using the user's resource usage history. According to our research, the work done till date is limited to other specific cloud providers or private servers but none related to Microsoft Azure. The classification algorithm that we use is LSTM. However, we have also report and document results obtained by ARIMA, SVM and Bayesian Networks. The best performance is given by LSTM.