Distributed Parallel File System Copy Of The Management Strategy

With the development of the Information Society, especially with the development of the Internet, more and more information is digitized which caused that data increased drastically. As a result, how to store and manage these data will become the focus concerned by people. In the 70's and 80's, because these data were stored in the expensive middle or mini computers and the system manager had to back up the data periodically, the enterprise cost a lot. On the other hand, once these machines were out of work, the services provided by these machines had to be interrupted and the enterprise suffered big loss. So a new way of storage is being researched. It is an important direction to store data in a distributed server system, which is composed by several high-performance PCs connected by the high-speed LAN and managed by the distributed operating system.Based on the Linux for the distributed parallel servers, DPFS (Distributed Parallel File System), one core of the distribute system, was designed to manage the data intelligently so as to get the highest performance and reliability for the system.The development overview and the new demands of data storage are described first. Then the aspect that the distributed file system affects data storage and the problem in the design of distributed file system are introduced. The goal we must achieve, when designing distributed parallel file system, was analyzed also. The thesis present the logic structure of DPFS, the structure and the flush strategy of the module of directory cache and its role in the read-write operation, duplicate table's physical and logical structure, the management and synchronization algorithm of the duplicate table, and the model and the management algorithm of the module of intelligent duplicate management. Also the effects that DPFS exerts to the system reliability and the read and write performance compared to EXT2 were analyzed. At last the further work of the distributed parallel file system was discussed.The redirect of open operation and the intelligent replication management were focused on. We put forward an intelligent duplicate management model including the methods of recording the duplicate information, the strategy of creating and deleting duplicate. Then we analyzed quantitatively the implement of the intelligent duplicate management algorithm by an example of the BOD (Broad band service On Demand) system.