The realization of high-performance cluster management system based on autonomic computing

Author

Jie, Wang ; Yu, Zeng

Author_Institution

Sch. of Manage., Capital Normal Univ., Beijing, China

fYear

2012

fDate

29-31 May 2012

Firstpage

2515

Lastpage

2518

Abstract

Cluster is now the main architecture of high-performance computers. We discuss distributed and hierarchical autonomic management mechanisms based on partition and election. Three autonomic management strategies are proposed, including a new definition of autonomic element, a new formula to calculate the number of nodes managed in a logical partition and a new trigger-bully election algorithm. Based on the three strategies, we design and realize a high-performance cluster management system ACView in this paper, and this ACView system is now successfully applied in a real high performance cluster.

Keywords

distributed processing; fault tolerant computing; ACView system; autonomic computing; autonomic element; hierarchical autonomic management mechanisms; high-performance cluster management system; trigger-bully election algorithm; Algorithm design and analysis; Clustering algorithms; Computational modeling; Computers; Monitoring; Nominations and elections; Partitioning algorithms; autonomic computing; election algorithm; high-performance cluster management; logical patrition;

fLanguage

English

Publisher

ieee

Conference_Titel

Fuzzy Systems and Knowledge Discovery (FSKD), 2012 9th International Conference on

Conference_Location

Sichuan

Print_ISBN

978-1-4673-0025-4

Type

conf

DOI

10.1109/FSKD.2012.6234057

Filename

6234057