CloudStreamMedia: A Cloud Assistant Global Video on Demand Leasing Scheme

Abstract

Cloud computing is a new computing paradigm that takes all resources as services, and it is not only agile, but also scalable. With the development of cloud computing, video on demand has become one of the most popular applications over the Internet. Currently, there is a trend of using cloud data centers and virtualization technologies to expand large-scale video streaming services with higher quality and lower expense. In this paper, we present CSM (Cloud Stream Media), a scheme that books the minimum resources from global data centers to match its demand and dynamically adjusts all resources to effectively meet the users' requests and guarantee a certain kind of quality of service, thus enhances the utilization and decreases the cost. CSM first predicts the stream media's future demand and data center's workload by using ARIMA model, and then performs a locality-aware resource booking (LARB) algorithm to lease the necessary resource from globalized cloud service providers in a long time. In order to handle prediction inaccuracy and the short-term demand peeks, CMS also introduces an inaccurate prediction handle strategy and performs auto scaling. We evaluate our scheme by combining both real world data and simulation. The results show good accuracy of our prediction and about 20% cut of total cost.