Research On Key Technologies For The Client Of Distributed File System

The Internet technology changes with each passing day, and penetrates into everyaspect of our lives unconsciously, at the same time produces massive data and the quantityis in rapid growth. Large-scale distributed clusters and distributed file systems arenecessary to handle and process the massive data.Distributed file system is the necessary tools for visiting and using of big data. As theentrance for users to use and manage the distributed file system, when using thedistributed file system, the client of distributed file system shows users the entire filesystem abstraction, shielding all kinds of complex interactions, communications andprocessing logic among the client, the metadata server and the data server. Whether theclient of a distributed file system is designed and implemented reasonable and perfect ornot, will greatly influence the overall file system performance and reliability.With the analysis of the clients of common distributed file systems, the client of theRaccoonFS is designed and implemented. Aiming at the problem of deep directorymetadata requests efficiency and the problem of single client’s speed of data operations,this paper put forward several key technologies to solve these problems with carefulanalysis and research. By adding client-side metadata cache metadata to optimizemetadata server’s name space query and by asynchronous reading/writing for data services,the client improved the whole performance and reliability of the RaccoonFS.Metadata operation tests show that the optimized RaccoonFS client could improvethe deep directory file metadata operations at least15%in the worst case, and RaccoonFSof metadata operating speed is superior to the HDFS. Single client’s writing speed of filedata is3times faster, and the speed of reading data increases at least30%. The aboveresults show that, the proposed RaccoonFS client key technologies of metadata and dataoperation performance optimization is significant.