Prediction-based resource allocation in clouds for media streaming applications

Media streaming applications have recently attracted large number of users in the Internet. With the advent of these bandwidth-intensive applications, it is difficult to provide streaming distribution with guaranteed QoS relying only on central resources at the content provider. Cloud computing offers an elastic infrastructure that media content providers (e.g., VoD provider) can use to obtain resources on-demand. Since a media content provider is charged for amount of resources (bandwidth) rented from the cloud, an open problem is to decide on the right amount of resources allocated in the cloud and their reservation time such that the financial cost on the content provider is minimized. We consider a practical pricing model that is based on a non-linear tariff (i.e., a pricing scheme that depends non-linearly on the resources purchased or time reserved). We formulate the optimization problem based on prediction of future streaming demand. We then propose a simple (easy to implement) algorithm for resource allocation that exploits the non-linearity in the price contract, while ensuring that sufficient resources is reserved in the cloud without incurring wastage. The results of our numerical evaluation and simulations show that the proposed algorithm mimics the optimum solution very well.