Abstract
Live media streaming has become one of the most important applications over the Internet. However, having an all-round scalable, reliable, responsive and cost-effective solution for delivery of live video over multiple platforms is still a challenge. To deliver live video over multiple platforms like laptops, mobile phones, tablets, gaming consoles, etc., the video needs to be encoded to appropriate format based on the device on which it is to be rendered. The traditional way of using hardware encoder has been known to be expensive and inefficient. Cloud provides virtually infinite on-demand resources where the software encoders can run and the tenants are charged only for the renting period. The main idea is to vary the number of encoding/transcoding and streaming servers dynamically based on the user demand for each type of stream. We propose a framework to deliver live video over multiple platforms and a neural network based prediction model to predict the video demand for each type of stream taking parameters like device type, region, etc. into consideration. The use of cloud for encoding and streaming video makes the overall implementation scalable and cost-effective.
