Design and implementation of two-level metadata server in small-scale cluster file system
The reliability and high performance of metadata service is crucial to the store architecture. A novel design of a two-level metadata server file system (TTMFS) is presented, which behaves high reliability and performance. The merits both centralized management and distributed management are considered simultaneously in our design. In this file system, the advanced-metadata server is responsible for manage directory metadata and whole namespace. The double-metadata server is responsible for maintaining file metadata. And this paper uses the Markov return model to analyze the reliability of the two-level metadata server. The experiment data indicates that the design can provide high throughput.
Key wordsreliability two-level metadata server Markov return model failover
CLC numberTP 316
Unable to display preview. Download preview PDF.
- Wang Zhaofu, Zhang Wenhao, Liu Zhong. A Reliability Analysis Model of the Meta-Data Server in LCFS [J].Computer Engineering and Science, 2005,27(5):54–58.Google Scholar
- Brandt S A, Miller E L, Long D D. E,et al. Efficient Meta-data Management in Large Distributed Storage Systems [C]//The 20th IEEE/11th NASA Goddard Conf on Mass Storage Systems and Technologies. San Diego, CA, April, 2003:290–298.Google Scholar
- PVFS. The Parallel Virtual File System [CP/OL]. [2005-12-30].http://parlweb.parl.clemson.edu/pvfs.Google Scholar
- Xiong Jin, Fan Zhihua, Ma Jie,et al. Metadata Consistency in DCFS2 [J].Journal of Computer Research on Development, 2005,42(6):250–267.Google Scholar
- Jeff B. NFS: Hunting for a Cross-Platform File System [J].Network Computing, 1998,9(12):101–104.Google Scholar