Design And Implementation Of Streaming Media Traffic Generator Based On Tilera Many-Core Platform

Streaming media traffic generator is the tool that generates quantitative streaming media network traffic in the network. It plays an indispensable role in the research of network architecture, new network protocol, network services, etc.According to its realization method, the existing streaming media traffic generator is divided into two categories: the method based on streaming media traffic model and the method based on streaming media user’s behavior model. Through the former generate streaming media traffic by structuring and sending packets, which compared with the latter, it still exists the generated rate of streaming media traffic and real problems. While the latter is better than the former in these aspects, its design is mainly based on C/S architecture and common hardware platform. Not only is the operation not convenient, but also the generated number of concurrent users is small. In addition, the rate of streaming media traffic is low and lacks of authenticity. Thus, the latter still can’t simulate such scene: a large number of users to simultaneously watch videos in the future, and the generated streaming media traffic is difficult to meet the demand of the gigabit or over experimental network.Aiming at the foregoing problems, this thesis has proposed a new system architecture of streaming media traffic generator based on Tilera many-core platform in B/S and distributed system architecture model, including Tilera many-core hardware platform and software architecture. The primary work is summarized as follows:(1) In order to ensure the authenticity of the media traffic flow, this thesis chooses streaming media user’s behavior model as the basis. By sending stream media protocol requests to streaming media servers in the Internet, tricking them generate large-scale traffic in the passing experiments network, thereby ensure the authenticity of the media traffic flow.(2) To improve the network throughput of the media traffic flow, this thesis adopts Tilera many-core platform as the hardware foundation of the system. Parallel tasks on Tilera many-core platform are assigned and socket non blocking mode are choosed in the system, thus, the network throughput of the media traffic flow is improved.(3) In order to increase the number of concurrent users of streaming media, this thesis designs a mechanism for keeping streaming session alive, and realizes the combination of thread pool and epoll technology on Tilera many-core platform, then the number of concurrent users of streaming media is increased.The experimental results show that the system have advantages of higher throughput rate, concurrent users and authenticity in different scenarios. This system can simulate 20000 streaming media concurrent users that watching videos online concurrently, which also shows that it has achieved the design goal in operability, scalability and stability. This thesis can provide a better reference value for the research and implementation of streaming media traffic generator.