Media Data Management Scheme For P2P Based Streaming Services

With the help of Peer-to-Peer computing, massive media streaming data can bedistributed among the users, which can decrease the overhead of servers greatly. So,Peer-to-Peer computing is the base to support large-scale network streaming services. But,based on P2P architecture, there are several challenges in the media data management:1)there are large-scale peers in the streaming system, and all peers will send the updateinformation of local media data periodically, which will bring big overhead to indexservers to deal with the update information. And the big update information also will leadthe long searching delay;2) network environment in P2P computing is heterogeneous andjitter. The system cannot send the media services with the same quality to all peers, whichare in different network environment. A media distributing scheme based on multi-rate canbe selected to solve the problem;3) peers in the P2P network are unstable, which willbring the big challenges to the media buffering. Totally, above challenges can be describedinto three topics: media indexing, media distributing and media buffering. The three topicsare very important to build large-scale media streaming services.About the media indexing problem in large-scale P2P based media streaming services,we have proposed a distributed media indexing structure and optimized searching schemesbased on periodical information sharing and exchanging. Distributed media data indexingstructure includes distributed index servers organization, distributed media indexinformation structure and periodical updates management from normal peers to indexservers. And searching scheme with high efficiency can do work by two strategies: one isdistributed searching scheme with the help of the agent of index manager; the other one isthat the searching request can be sent to anyone index server because all index servers willshare index information periodically. Based on the above schemes and algorithms,compared with central indexing scheme, the experiments prove that distributed indexingscheme has the same searching ability, but performance good at running overhead andsearching delay.About the media data adaptive transmission problem in large-scale P2P based mediastreaming services, we have proposed an adaptive media data distribution scheme in unstable network environment based on multi-rate topologies. It is a new transmissionstrategy with mixed-rate, which can dynamically replace the streaming rate according tothe network environment’s change. The main algorithms and functions in the schemeinclude: buffer management of multi-rate media data, layered topology design ofmulti-rate media data, scheduling of multi-rate media data and streaming rates switchingdecision. Based on the above schemes and algorithms, compared with traditional schemebased on single streaming rate, the experiments prove that it has good startup delay,improved playback continuity and decreased overhead of media servers, even though thebandwidth occupations of peers have been increased a little, which is acceptable. AndCompared with multi-source streaming distribution scheme, it has better startup delay,server load and almost the same playback continuity.About the media data buffering and sharing problem in large-scale P2P based mediastreaming services, we have proposed an efficient buffering and balanced sharing schemebased on multiple integrated topologies. The new scheme includes media data clip’sorganization based on time, states maintaince and distribution algorithm of buffer,balanced media data distribution optimization algorithm in the distributed topology. Basedon the above schemes and algorithms, the experiments prove that the new scheme hasgood performance. With the help of the buffering scheme, at least50%overhead of mediaservers have been transferred into the distributed peers. The new scheme can improve theefficiency of media sharing in the whole network, and help peers against the networktremble. The performances of the new scheme are better than those of single topologybased buffering schemes.