Research On Data Synchronization Mechanism Of The Information Repository Federation In Publish/Subscribe Model

In a distributed system, Publish/Subscribe model, as a communication mode ofdata-driven, because of its advantages of simple, high-speed and so on, is applied in a largenumber of critical systems. To establish a connection between the communication entities inthe system is running, information warehouse node distributed, responsible for an importantmission. During the running of system, information repository node shoulders the importantmission in establishing connections between distributed communication entities. This caneasily make the information repository become the bottleneck of the performance of theentire system, once the information repository node failure or unable to connect due tonetwork congestion or other reasons, the entire Publication/Subscription model system willnot be able to go on running. To solve this problem, the method of adding redundant nodes toenhance the robustness and stability of the system, is wildly used, which leads to datasynchronization problems between information repository nodes. However, researches andapplications about this problem is still relatively preliminary at present, in most of theexisting system, ensuring reliable network connection between multiple informationrepository nodes are required, if there is disconnection or other problems, during the runningtime, the broken line information repository nodes will not be able to re-access the system.When the system running in the poor network environment with few redundant nodes, theproblem described above will not be able to guarantee of long-term stable running of system.In order to solve the above problem, this paper, proposed a data synchronization modelbetween information repository nodes which applies in a Publish/Subscribe model, bystudying the characteristics of the existing Publish/Subscribe model system, combining withtraditional data synchronization technology, to further strengthen the robustness and stabilityof entire system. By building the information repository federation, we will have effectivemanagement about the structure of the multiple information repository nodes; Byconstructing the data synchronization model of the Information Repository Federation, wewill have the effective protection about the data consistency of the multiple informationrepository nodes. Meanwhile, in order to avoid the dependence of the data synchronizationprocess on the control node of the fixed center, by introducing data synchronization central node coordination strategies to control peer data synchronization model completely; In orderto take into account the efficient data distribution characteristics of the Publish/Subscribemode, introducing the synchronization mechanism based on message queue and theapplication of state table to maximize data synchronization execution efficiency.At the end of this paper, we proved the correctness of the Information RepositoryFederation’s data synchronization model the verification experiment; Through comparativeexperiments, we analyzed the performance impacts of the Publish/Subscribe model systemcaused by data synchronization model. Finally, we analyzed the experimental results andillustrated that the data synchronization model of the Information Repository Federal haspractical significance.