Live streaming of user generated videos: Workload characterization and content delivery architectures

摘要

We are recently observing an increase in popularity of Web 2.0 applications offering users the possibility of transmitting their own live streaming media content. The considerable amount of network bandwidth required by such applications, especially at peak usage, poses significant challenges to their scalability and popularity. In this context, we propose and evaluate alternative content distribution solutions that contribute to improve the scalability of this type of system. This work encompasses two major steps. First, we collect data from a popular online live video sharing service, and perform a characterization of key aspects of its workload. Next, driven by some of our findings, we propose two variations of a new hybrid content delivery architecture, which combine elements from the traditional centralized client-server and Peer-to-Peer architectures. Our characterization findings are also used to drive the design of a system simulator, used to evaluate the proposed strategies and the traditional architectures, considering different scenarios as well as metrics capturing the interests of both users and service providers. Our results indicate that the hybrid strategies yield the best tradeoffs between quality of service perceived by users and server resources consumption.