In Method We Trust: Towards an Open Method Kit for Characterizing Spot Cloud Service Pricing
Zheng Li, Xuefei Li, Bing Li · 2019
Based on market-driven mechanisms that can improve utilization of idle compute resources at dynamic prices, spot cloud services are becoming increasingly popular to reach a win-win situation of providers' revenue maximization and consumers' budget optimization. Nevertheless, unlike fixed pricing schemes, the spot pricing scheme is both psychologically and practically sophisticated for people to understand and employ. As such, characterizing spot cloud pricing has been identified to be crucial and beneficial for various purposes ranging from facilitating service procurement to addressing service interruptions. In addition, it is also noteworthy that the spot cloud market is inherently volatile, and the providers' (e.g., Amazon's) pricing policies can change from time to time. Consequently, the previous analysis results can quickly be out of date, and the previous analysis methods can barely be reusable if their details are not specified. Therefore, we decided to develop a domain-specific method kit and make it open to improve the repeatability, replicability, and reproducibility of spot pricing characterization studies. This paper reports the typical content of this method kit, including a group of central tendency analysis methods and two types of distribution modeling analysis methods. In a generic sense, we particularly argue that open methods act as a higher-level strategy over open-source tools and open-access data for scientific studies in any research domain.