Pageserver: High-Performance SSD-Based Checkpointing of Transactional Distributed Memory

Author

Gerhold, Steffen ; Kaemmer, Nico ; Weggerle, Alexander ; Himpel, Christian ; Schulthess, Peter

Author_Institution

Inst. of Distrib. Syst., Ulm Univ., Ulm, Germany

Volume

1

fYear

2010

fDate

19-21 March 2010

Firstpage

235

Lastpage

239

Abstract

The Rainbow cluster operating system provides fault tolerance by storing incremental checkpoints on a solid state disk (SSD) drive. As the access characteristics of SSD drives differ substantially from those of traditional magnetic hard disk drives (HDDs) new interesting techniques of storing checkpoints in an elegant object-oriented fashion can be applied. The usage of transactional distributed memory (TDM) simplifies the distributed checkpointing algorithm as it automatically guarantees consistent checkpoints of the entire Rainbow cluster without additional coordination overhead.

Keywords

checkpointing; disc drives; distributed memory systems; fault tolerant computing; hard discs; operating systems (computers); SSD; distributed checkpointing algorithm; fault tolerance; magnetic hard disk drives; rainbow cluster operating system; solid state disk drive; transactional distributed memory; Checkpointing; Clustering algorithms; Computer applications; Distributed computing; Drives; Hard disks; Java; Operating systems; Throughput; Time division multiplexing;

fLanguage

English

Publisher

ieee

Conference_Titel

Computer Engineering and Applications (ICCEA), 2010 Second International Conference on

Conference_Location

Bali Island

Print_ISBN

978-1-4244-6079-3

Electronic_ISBN

978-1-4244-6080-9

Type

conf

DOI

10.1109/ICCEA.2010.53

Filename

5445831