Multi-Cloud simulation environment for WebRTC streams

Abstract: Real-time streaming is becoming popular nowadays on the internet. WebRTC is a promising web technology that enables media stream transmission between two browsers without other third-party plugins. However, traditional WebRTC applications can only establish peer-to-peer (P2P) communications, which cannot be directly used in one-to-more streaming scenarios such as a multi-party video conference. This thesis project presents a development of a multi-cloud simulation framework to implement software multicast of WebRTC streams to enable oneto-more real-time streaming. The framework can generate a cloud network topology with a simple script, and provides flexible ways to model the network links with parameters such as bandwidth, packet loss, and link latency. By using selective forwarding units (SFUs), a stream publisher only needs to send a single copy of the data to the cloud, the data packets are duplicated and forwarded to multiple subscribers. In addition, three resource allocation algorithms are included in the framework to determine the data center for a task. To evaluate the performance, this framework enables people to monitor the throughputs and running streams on the data centers during the experiments. We develop benchmark applications as substitutes for real WebRTC traffic. These applications can generate UDP stream packets with the same dimension of WebRTC packets and provide the customization of stream parameters. In this report, we present a comparison of the stream performances under different allocation algorithms. Finally, as an outcome of this project, we design an integrated functional test to simulate a real-world scenario. The result shows that the framework is able to work well on complex real cases and simulate most of the multi-cloud networks.