Pricing the Cloud: An Adaptive Brokerage for Cloud Computing

Using a multi-agent social simulation model to predict the behavior of cloud computing markets, Rogers & Cliff (R&C) demonstrated the existence of a profitable cloud brokerage capable of benefitting cloud providers and cloud consumers alike. Functionally similar to financial market brokers, the cloud broker matches provider supply with consumer demand. This is achieved through options, a type of derivatives contract that enables consumers to purchase the option, but not the obligation, of later purchasing the underlying asset—a cloud computing virtual machine instance—for an agreed fixed price. This model benefits all parties: experiencing more predictable demand, cloud providers can better optimize their workflow to minimize costs; cloud users access cheaper rates offered by brokers; and cloud brokers generate profit from charging fees. Here, we replicate and extend the simulation model of R&C using CReST—an opensource, discrete event, cloud data center simulation modeling platform developed at the University of Bristol. Sensitivity analysis reveals fragility in R&C’s model. We address this by introducing a novel method of autonomous adaptive thresholding (AAT) that enables brokers to adapt to market conditions without requiring a priori domain knowledge. Simulation results demonstrate AAT’s robustness, outperforming the fixed brokerage model of R&C under a variety of market conditions. We believe this could have practical significance in the real-world market for cloud computing. Keywords—CReST; simulation; cloud computing; brokerage