The Research Of Channel Switching And Resource Balance In Multi-channel P2P Streaming System

With the development and prevalence of the Internet, P2P streaming media has attracted people's close attention in the past years. Streaming media has caused a profound impact on people's work and lives with its abundant contents and its special feature of being real time. P2P streaming media software, such as Coolstreaming, PPlive and PPstream, has never been so popular in recent years, and drew people's unprecedented attention.Up to now, the system of streaming media provides hundreds of channels. However, this system has bunch of performance problems, including long time delay when switching channels, bad performance of rare-used channels. All these originate from channel stir caused by channel switching and imbalance of resources among channels.When a node is switching channels, this node releases the channel, which means breaking the connection to system, resulting in node's failure. The neighboring nodes of this node in the system will suddenly lose the uplink bandwidth supplied by it and have to look for new nodes which have necessary data block and sufficient uplink bandwidth. This will cause the time delay or even loss of data when the neighboring nodes receive data, and further lead to uncontinuity of services. In addition to this, nodes have to look for new nodes in the new channel to download video data, and this can only be achieved by keeping communicating with nodes enlisted in the list downloaded from server, also increasing the time delay.In future, we forecast that a quantity of user-made temporary channels created based on network camera or wireless hand-held equipment, will emerge. The number of these small channels could be thousands. Each of these channels which get data from a relative low-speed link, such as a wireless PDA, could only have 10 to 1000 viewers. For the difference of the number of audience, channels those have less audience have worse performance generally. In contrast, channels with more audience have wider upload bandwidth, plus, the more people, the easier to find closer nodes except for faster transmission speed, which further brings about better fluency. From above, we can get a conclusion that imbalance of allocation of resource between hot channels and rare-used channels, causes performance difference.The main contents of this paper: ①When peer switches the channel, leading to the lose the link with the neighbor peers, resulting the channel churn, we proposed an elastic switching mechanism based on the slow reaction.②In order to balance resource among channels, this essay raises a method named Balanced Group. This essay demonstrates the construction of balanced group, distribution and recycling mechanism. For the purpose of making resource allocated more efficient, a method of predicting the number of nodes is presented. Furthermore, a method of dividing main stream into sub-stream to minimize the overall burden of nodes in a balanced group allocated to channel is also mentioned. This method can balance resources among channels, providing bandwidth for channels needing resources and using idle resource reasonably, to reduce the pressure taken by servers.③Design and implement multi-channel P2P streaming system simulator for simulation. Simulation results show that the elastic switching mechanism based on the slow reaction and balance group are very effective.