Research And Implementation Of Massive File Organization Based On Attributes And Links

The rapid development of Internet technology has resulted in the emergence ofvarious types of data resources. It is difficult for user to analyze and layering theinformation structure they faced, so how to organize files effectively has become to be aserious problem. Most of current file systems are mainly used to solve the performanceand reliability problems in data storage filed, and the content of the data itself and thesemantic information, including the data characteristics, between files was less concerned.To solve the problems above, the academic brings up the semantic file system. Theyextended the semantic relationships between files by introducing a new model in existingfile system, but the file system structure was not modified completely. And queryperformance was not considered to be an important factor. Therefore, it is very meaningfulto design an efficient file management system.We analyze and discuss the three typical information organization mechanism andfile retrieval mechanism, and summed up the features and shortcomings in the work of fileorganization and management in current file systems. At last, we design and implement aprototype system that based on extended attributes and the internal links. The files areorganized as a flat file structure in the prototype system logically. We provide amechanism that compatible with the POSIX specification to achieve access by virtual pathand a flexible interface for the extended attributes and internal links operation. In thesystem prototype, extended attributes are used to ensure the integrity and scalability offile’s semantic information, and internal links, which can avoid the traditional file systemassociation failures, are used to achieve the automatic management of file association. Wedesign the personalized logical view to help user to location the file needed rapidly andexactly, and use inverted file indexing mechanism to improve the performance fileretrieval.With the functional tests, the prototype system proved us that it has implemented thefunction of metadata extensions, file associations and file retrieval etc. The systemperformance is tested too. The results show that it is faster than third-party search tools inthe document retrieval, and the performance is improved from48%to64%. For the operations of adding extended attributes and some other meta-data operations, theresponse time of prototype system is generally about20microseconds. As the realizationtaking into account, a user space file system framework is used in prototype system, whichleads to almost20%performance reduction in the read and write operations test,compared to kernel mode file system.