A Novel Cluster Parallel File System

Author

Wei, Wenguo ; Dong, Shoubin ; Zhang, Ling ; Li, Jialin

Author_Institution

Guangdong Key Lab. of Comput. Network, South China Univ. of Technol., Guangzhou

fYear

0

fDate

0-0 0

Firstpage

119

Lastpage

123

Abstract

We propose a novel cluster parallel file system: high-throughput & scalable parallel file system (HTSPFS). There are some new features in it such as automatic adaptive file striping policy based on file size, request arrival rate, concurrent of accessing files, storage servers´ networking and disk performance; prefetching read-request and writing behind write-request to enable a higher throughput in HTSPFS. We give a detail discuss of Markov spatial predicting method for read request, experiment results show that Markov predicting technology has high prediction veracity. Overall, HTSPFS has some autonomies characteristics

Keywords

Markov processes; concurrency control; file organisation; parallel processing; workstation clusters; Markov predicting technology; Markov spatial predicting method; automatic adaptive file striping policy; cluster parallel file system; high-throughput file system; prediction veracity; scalable parallel file system; Aggregates; Computer networks; Computer science education; Concurrent computing; Distributed computing; File systems; Laboratories; Network servers; Prefetching; Throughput;

fLanguage

English

Publisher

ieee

Conference_Titel

Database and Expert Systems Applications, 2006. DEXA '06. 17th International Workshop on

Conference_Location

Krakow

ISSN

1529-4188

Print_ISBN

0-7695-2641-1

Type

conf

DOI

10.1109/DEXA.2006.14

Filename

1698319