Cloud Cost Forecasting

Abstract

Although scalability has been one of cloud computing’ greatest merits, it leaves the consumer and the operator with uncertainty about future costs and corresponding capacity requirements. The IT industry is partially transitioning towards IT as a utility, which impacts both price and costs (Truong Huu et al. 2010). This uncertainty could prove to be too difficult to conquer for some organizations and holds back acceptance of cloud computing technology. Several cloud infrastructure providers have recently introduced cloud cost calculators of some sort to come up with an estimation of monthly costs (Corporation 2012; IBM corporation 2012). An important assumption made by these cost calculators is that customers are required to know how much they will consume, which. Until now there has not been a significant contribution in literature that addresses the forecasting of costs for cloud services. This research adopts a design science research strategy (Hevner 2007). An IT artifact is created that is able to forecast cloud costs. The current state-of-the-art in the understanding of cloud costs in published literature is researched using a structured literature review (Kitchenham 2007). The majority of researchers include cost savings as a rationale for their research. Several quantitative forecasting techniques have been researched according to the classification of Armstrong (Armstrong and Green 2005). A decision matrix was created to aid the selection of the proper forecasting technique. A technical solution is created that forecasts infrastructure level usage on CPU, memory, network and storage which combined create a significant portion of the cloud costs. A controlled experiment is used to test the application of forecasting theory in the context of cloud costs. An expert review with five cross domain experts is performed to validate – and help interpret – research findings. After having conducted this research, it is concluded that extrapolation forecasting on cloud costs has great potential for cloud platform operators. In this work I present a literature review on cloud costs, a comparison of the application of forecasting techniques in the context of cloud computing costs, a controlled experiment with a forecasting technique on a real-life case, a validation by means of an expert review, and research directions for future research.