A financial brokerage model for cloud computing

One of the major benefits of cloud computing is the ability for users to access resources on a pay-as-you go basis, thereby potentially reducing their costs and enabling them to scale applications rapidly. However, this approach does not necessarily benefit the provider. Providers have the responsibility of ensuring that they have the physical infrastructure to meet their users' demand and that their performance meets agreed service level agreements. Without an accurate view of future demand, planning for variable costs such as staff, replacement servers or coolers, and electricity supplies, can all be very difficult, and optimising the distribution of virtual machines presents a major challenge. Here, we explore an extension of an approach first proposed in a theoretical study by Wu, Zhang, the resources can then subsequently be provided to clients who demand it. We implement an extension of the WZH model in an agent-based simulation, using asset classes and price-levels directly modelled on currently available real-world data from markets relevant to cloud computing, for both service-providers provisioning and customers' demand patterns. We show that the broker profits in all market conditions simulated, and can increase her profit by up to 36% by considering past performance when deciding to invest in reserved instances. Furthermore, we show that the broker can increase profits by up to 33% by investing in 36-month instances over 12-month. By considering past performance and investing in longer term reserved instances, the broker can increase her profit by up to 44% for the same market conditions.