Attribute Management And Metadata Pre-fetching Research On The Windows Client In Object-Based Storage System

Object-Based Storage System is constituted by Clients, Metadata Servers and Object Storage Devices. The MetaData Servers store and manage the metadata centralized, and the Object Storage Devices have the real data and objects. Division of the management and storage can provide more methods to improve the whole performance with the help from intellectual devices. The clients are the very crucial part of the system. A high qualified, stable and friendly client file system can result in a better performance of the system.In the design of the Window Client File System, this thesis uses Installable File System Driver Development Kit, IFS DDK from Microsoft. In the file system, the requirements for file attributes can be served quickly, and serve the users very steadily.In the Windows File System, almost every operation about files and objects comes along with requirement for metadata. In order to alleviate the significant time cost of network and load on server, pre-fetching files to cache memory before they are used remains a promising way to mitigate the problem. By studying and testing the special characters of metadata, this thesis designed a self-adaptive metadata pre-fetching model, which was implemented by learning from the users' or programs' access pattern, and discovering the relationship between the data, then fetched the ones which will be needed in the near future to the client, so that increases the accessing hit ratio.At the end, this thesis made some simulation and proves that this method bring a better performance of the whole storage system, increasing about 10% Cache hit rate, and saving about 20% average responding time.