Research On A Data-Driven Media Streaming Application Layer Multicast Strategy

Nowadays media streaming technology is applied to many important fields, such as digital library, remote education, on-demand video, real-time IP TV and cooperative communication. At present distributing media streaming to end systems over the Internet becomes a hot research in Computer Science field. Traditional Client/Server model isn't able to adapt to large-scale distribution of media streaming because of Servers'performance bottleneck. In order to support large-scale data distribution IP multicast has been presented for more than decades. However, IP multicast hasn't been deployed in large by far. Because its deployment has to modify basic facilities of Networks and there is much intrinsic confinement in IP multicast. In recent years, aiming at this problem researchers have proposed application layer multicast strategy (ALM) which transfers multicast function onto application layer. ALM is more flexible than IP multicast and easier to deploy and implement.Performance Index of ALM includes transmission delay, bandwidth utilization ratio, load distribution, robustness, etc. Complexity of implementing strategy itself should also be considered. All of typical ALM strategies try their best to make trade-off among these performances. Some strategies make a suitable trade-off by leveraging complex multicast communication protocol. However, they are hard to implement. This paper designs a data-driven media streaming Application layer multicast strategy (DSALM), it organizes nodes into an unstructured Overlay Network by a distributed, Self-Organizing mode. Every node fetches data from multiple nodes at one time and optimizes data dynamically. Scheduling algorithm which is core of the strategy is data-central and executed by receivers. It makes a better trade-off among these performances. The features of DSALM mainly includes: (1)Easy to implement, it doesn't have to construct and maintain a complex multicast tree, and also doesn't have to deploy multicast communication protocol. (2) Efficient,end systems can join in multicast group quickly, and have a good performance of data transmission delay. (3) Fair,DSALM tries its best to distribute the data-forwarding load among all of nodes. This is compatible with P2P idea and also behaves the fairness of data-forwarding load distribution. (4) Robust,in high dynamic Networks, it can still ensure Service Quality of media streaming of most of nodes.DSALM and HMTP (a typical ALM based on multicast tree) are emulated in the simulated Networks environment. The experimental data shows:DSALM is better...