Usage Patterns and the Economics of the Public Cloud.

We examine the economics of demand and supply in cloud computing. The public cloud offers three main benefits to firms: 1) utilization can be scaled up or down easily; 2) capital expenditure (on-premises servers) can be converted to operating expenses, with the capital incurred by a specialist; 3) software can be ``pay-as-you-go.'' These benefits increase with the firm's ability to dynamically scale resource utilization and thus point to the need for dynamic prices to shape demand to the (short-run) fixed datacenter supply. Detailed utilization analysis reveals the large swings in utilization at the hourly, daily or weekly level are very rare at the customer level and non-existent at the datacenter level. Furthermore, few customers show volatility patterns that are excessively correlated with the market. These results explain why fixed prices currently prevail despite the seeming need for time-varying dynamics. Examining the actual CPU utilization provides a lens into the future. Here utilization varies by order half the datacenter capacity, but most firms are not dynamically scaling their assigned resources at-present to take advantage of these changes. If these gains are realized, demand fluctuations would be on par with the three classic industries where dynamic pricing is important (hotels, electricity, airlines) and dynamic prices would be essential for efficiency.