Dynamic resource allocation for video transcoding with QoS guaranteeing in cloud-based DASH system

Due to diverse network conditions and heterogeneous devices, there may be various video demands with different video qualities and formats from the client side. Compared to keeping all necessary copies for the same video, video transcoding in real-time should be an essential solution. The complex nature of video transcoding enables cloud computing to be uniquely suitable for dynamically providing transcoding resource. However, due to the fluctuation and uncertainty of the future transcoding demand, it is still a challenge to dynamically determine the optimal resource allocation to save cost while guaranteeing the Quality of Service (QoS). Overload may result in the transcoding jitter and increase the lateness which directly affects video freezes while over-provisioning naturally increases the cost. To address this problem, in this paper, by defining the transcoding jitter probability as a metric of QoS, we proposed a dynamic resource allocation algorithm based on the large deviation principle, which is capable of proactive calculating the optimal number of transcoding nodes for the upcoming transcoding demand subject to the transcoding jitter probability below a desired threshold. Finally, the experiments are performed on a cloud-based prototype system to show the attainable performance of the proposed resource allocation algorithm and verify that the proposed algorithm can make a good tradeoff between cost saving and QoS guaranteeing.