A new live video streaming approach based on Amazon S3 pricing model

Video has become a mainstream media source on the web, and live video streaming is growing as a prominent player in the modern marketplace for both businesses and individuals. Due to elasticity and on-demand nature that cloud computing provides, more and more video service providers (VSP) utilize cloud infrastructures to provide cloud-based video applications. In this work, we propose Cloud Live Video Streaming (CLVS) - a new approach to efficiently stream live video that is based on Amazon S3 pricing model. In the design of CLVS, when a source video is being recorded by a mobile device, on which then it is segmented and encoded. Next, those video segments are pushed into a designated Amazon S3 bucket. On end-user devices, the client program of CLVS directly retrieves the most recent video segment from the S3 bucket, then performs decoding and video playing back. By using Amazon S3 service, our CLVS employs what is referred to as a “serverless” design by eliminating the needs for an intermediary and persistently running streaming server. Thus, VSPs do not have to pay for the idle time found on the traditional streaming servers, but still are able to continually provide live video streams to end viewers. We implement a prototype of CLVS and optimize its performance by using multithreaded prefetching and caching techniques. In experiments, we compare our CLVS with an existing video streaming software - Wowza Streaming Engine. Experiment results indicate that CLVS not only outperforms Wowza in many aspects of video streaming performance, but also financially costs less.

I. Introduction

The growth of the streaming video industry has risen from $3.1 billion in 2013 to$4 billion in 2014 to $5.1 billion in 2015, and predicted to be$6.7 billion in 2016 [1]. A large number of businesses have recently explored live streaming. According to a recent survey, in the past 12 months 44% of businesses have conducted one or more live streaming video events, 35% do not plan to test or conduct a live streaming event, and 20% plan to test one or more live streaming events [2].