Cost-Aware On-Demand Resource Provisioning in Clouds

Cloud computing has been gaining increased popularity among companies for a couple of years. Low upfront costs and the enormous elasticity make it an attractive alternative to in-house IT solutions. However, most companies still tend to overprovision their IT infrastructures induced by the buffering behavior to avoid violations of service level agreements. The introduction of dynamic scaling in IT infrastructures promises enormous cost-saving potential. We identify factors of influence for cost-aware operation of an on-demand provisioned system and propose a novel provisioning model that attempts to optimally scale the cloud at any time. Following the design science paradigm we formulize the elements of our simulation framework and carry out extensive simulations under fine-grained real world settings. Compared to common operation strategies, our approach delivers superior performance results. Finally, we derive managerial implications based on the cloud customer's preference between cost-awareness and SLA compliance.