A simulation analysis of different allocation and pricing policies for cloud computing service providers

Official URL: http://risc01.sabanciuniv.edu/record=b1655745 (Table of Contents)

The cloud computing is regarded as a paradigm shift in nowadays IT world. As services of cloud computing behave like perishable products, revenue management techniques can be applied to increase cloud service provider's total revenue. In this thesis, we develop various methods for pricing and capacity allocation. We consider three types of instances; subscription, on-demand and spot instances. We introduce three allocation and pricing policies and propose 8 models based on their combinations. First, we establish a queuing mechanism for on-demand instances which are rejected initially by the cloud with a price incentive. Second, we consider an auction based model for spot instances and introduce two types of threshold policies where it is constant or dependent on the remaining capacity. Finally, the criterion for spot instances selection is based on expected revenue or bid of that customer. We simulate these models on several datasets and evaluate the models with di erent capacities. The results we obtain indicate the sensitivity of revenue based on the policies we propose over the studied datasets.