Research Of Metadata Management And Replication Algorithm For Massive Data

Because large scale DVE(distributed virtual environment) is not only required to simulate large scope geography environment and interaction of a great many entities, but also required to simulate the geography environment vividly and the behavior of complex entities, DVE needs to manage massive data objects, which spread all over the nodes of the wide area networks.At present, database system, file system and mixture of multiple systems are used to store and manage data objects in a great number of DVE systems. With the scale of DVE expanding, many problems such as the complexity of data management, the low speed of data access and the disunity of the interface for data access are appeared. Particularly, the architecture of these data management systems cannot be combined with the DVE system naturally. To thoroughly study the data management of DVE is crucial to the development of DVE.Based on the analysis of data and access pattern of DVE, an architecture of DVE based on grid and a hiberarchy architecture of massive data management system are proposed. A metadata catalog system, which includes center metadata catalog and local metadata catalogs, are designed and implemented. In this system, local metadata catalogs manage the metadata of the local nodes, and center metadata catalog manages all local metadata catalogs.To fulfill the requirement of data access efficiency in DVE, the prediction-based parallel replication strategy is proposed. This strategy includes parallel replication strategy and prediction strategy. The parallel replication strategy replicates from multiple replicas synchronously, it accelerates the speed of remote data access using the redundant paths of the IP networks. The prediction strategy replicates the data in the local nodes before data access, it optimizes the efficiency of data access by utilizing the unoccupied bandwidth of the networks. The results of simulation tests indicate that the prediction-based parallel replication strategy can reduce the delay of remote data access effectively, and it can improve the efficiency of data access.