Spot instance similarity and substitution effect in cloud spot market

Abstract

Customers in cloud spot market choose from a set of computing resources (spot instances) some of which are same along one or more of the dimensions of hardware configuration, hardware capacity, software, and location (or zones within the same region). While prior research in IS has shown that cloud market consumers do not substitute identical spot instances across distant locations or regions due to costs associated with latency, there is a paucity of research examining spot price dynamics of similar instances which are same along only some of the dimensions (or characteristics) within a region. Using the economic theory of substitution in production and consumption as our theoretical lens, our empirical analysis finds evidence that consumers substitute similar computing resources in the spot market. Further, we find that while hardware configuration similarity between two instances is positively related to spot instance substitution, this substitution effect along hardware configuration is moderated by similarity along hardware capacity and location (zone). Our work contributes to the literature on cloud spot market by providing empirical evidence of the substitution effect among similar spot resources within a region, and thus help in understanding the choices of consumers in this market. We conclude by discussing how practitioners can leverage these findings in the procurement of computing resources from cloud spot market.