Research On Replication And Placement Policy For Streaming Media

Currently, one of the hotspot issues on streaming media is the control of quality of service(QoS), which leads to the low enjoyment degree directly. The replication and placement policy isconsidered as an efficient solution. For different applications, the existing replication andplacement policies mainly include: replication and placement policies for cluster system,replication and placement policies for content distribution networks (CDN), replication andplacement policies for peer-to-peer (P2P) networks. However, there are still four problems:1)Replication and placement policies for cluster system model the system by assuming one ormore elements and results in instability of performance.2) Replication and placement policiesfor CDN did not resolve the territorial and dynamic characteristics of quests, and leaded to lowhit rate.3) There are low QoS of unpopular program as well as high Channel startup time andchannel zapping delay in internet protocol television (IPTV) system based P2P.The problems of replication and placement were discussed in this dissertation. The chiefwork of this article is as follows.1）A multi-objective evolutionary replication and placement policy was proposed tominimize the load imbalance degree and request blocking probability in a method of placementon the basis of the past populatity. Firstly, in the proposed policy, the issue was departed intoproblem of replication and problem of placement. The problem of replication was regarded as anissue of congressional apportionment, and an optimal replication policy was proposed forplacement, which copied media in a heuristic way, which always choose the one with biggestvalue of Q. Based on this, the problem of placement was regarded as an issue of multi-objectiveprogramming. A multi-objective evolutionary placement policy was proposed, and the resolutionwas received in a way of multi-objective evolutionary algorithm. A comparative simulationdemonstrated that the proposed policy can provide good load balancing and lower requestblocking probability.2）A dynamic replication and placement policy based on cognition was proposed to improvehit rate and reduce delay in CDN. In the proposed policy, a model was formulated to forecastpopularity of the contents referring to memory mode firstly. And then, a dynamic replication andplacement policy based on cognition was proposed. Files were copied and distributed tosurrogate servers through sensing, recording, deciding and learning. Meanwhile, messages oferror were used to impove the policy to obtain better resolution. The simulation results indicatedthat the proposed policy can reduce the cost of data transmission in CDN effectively as well assatisfy the variable demand of clients with low cost.3）A hybrid replication and placement policy for time-shifted IPTV system was proposed toimprove success rate. Firstly, the character of data of IPTV and the degree distribution werediscussed. It concluded that low channel startup time and channel zapping delay were achievedby placing I flames in peers with higher degree distribution. This policy placed the replica ofstreaming data in two manners: flash data mode and outdated data mode. Popular I flames of flash data were placed in cache with high degree distribution in flash data mode. Meanwhile, inoutdated data mode, outdated data were placed in a way of valve. The simulation resultsindicated that the proposed policy can improve the success rate effectively with little storagespace. Thereby, channel startup time and channel zapping delay got lower.