MFS: A Lightweight Block-Level Local Mirror of Remote File System.

JSW(2013)

引用 3|浏览48
暂无评分
摘要
Nowadays individual users often have more than one computing platform, such as traditional desktop in office, laptop computers in home, and mobile smartphones in outdoor. More and more users store their data into remote file system and access them over network in every time, but user has to download the whole file to the user computer before user wants to access one file, so user has to have the poor experience, especially access large file over wireless network. We have developed MFS, which is a lightweight client-side local mirror of remote file system. MFS mainly provides four mechanisms to solve the above problems. One mechanism is that MFS uses client-side file system based on disk as a persistent cache for files, and the capacity of the persistent cache is limited. The second mechanism is that MFS uses block-level granularity as the smallest unit of file access operations and transmission. The third mechanism is that taking event publish-subscribe pattern to keep files system consistent between user client and remote network file system server. The fourth mechanism is that taking different file consistency priority strategies for different types of files. All files will be stored on cloud or remote file system, but only some files which are often accessed recently will be stored on user client-side persistent disk transparently. So user can have a larger logical storage space than user local disk, and user also gets high accessing speed of accessing remote file system, which speed is close to the speed of accessing local disk file system. User's applications can always access files, and do not wait until all the blocks of the file is downloaded. Our evaluation demonstrates that MFS has a good performance, reliability, transparent scalability and simplicity. MFS can run on a diversity of user computers, and it is independent of any computer. © 2013 ACADEMY PUBLISHER.
更多
查看译文
关键词
block-level,file consistency,file system,persistent cache,publish-subscribe pattern,publish subscribe pattern
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要