On the improvement of uncertain cloud service capacity

Cloud service firms often adopt a hybrid strategy in which computing capacity comes from a combination of their privately owned infrastructure and a public infrastructure provided by vendors. This combination can be considered a cloud computing supply chain. A major complexity for firms in managing such a supply chain is that computing demand and private computing capacity are both random variables. This randomness may require public cloud computing to fill the supply–demand gap and lead to additional costs. In this paper we study a cloud service firms’ problem of determining the amount of private computing capacity that minimizes their expected total capacity cost. We first demonstrate that the optimal amount of private computing capacity is unaffected by its randomness. We then establish that the reduction in expected total capacity cost, due to improvements in the mean or the variability of private computing capacity, exhibit diminishing returns, which allows characterization of parameter impacts and the optimal improvement level. We also establish that when setting their computing capacity price, public vendors do not need to know the private computing capacity of cloud service firms using a hybrid strategy. Our research results should be helpful to cloud service firms making procurement decisions regarding private computing capacity and/or instituting computing capacity improvement programs. Our research results should also be useful to public vendors for pricing.