Critical issues in the design of a fault-tolerant multiprocessor database server

Author

Hvasshovd, Svein O. ; Sæter, Tore ; Torbjørnsen, Øystein

Author_Institution

Database Res. Group, Sintef Delab, Trondheim, Norway

fYear

1991

fDate

26-27 Sep 1991

Firstpage

226

Lastpage

231

Abstract

HypRa is a database server designed to meet extreme requirements from e.g. telecom applications. Fault-tolerance has been designed into HypRa by: (1) taking it into account at all hard-ware and software levels, (2) using a shared-nothing, coarse grained hardware platform with homogeneous nodes, basic fault detection, and a multiway interconnection network with dynamic rerouting, (3) doing fault masking and repair in software to achieve flexibility and dynamic reconfigurability, (4) using data fragmentation, fragment replication and replica allocation, and dynamic reconfiguration to automatically reproduce the data availability fault-tolerance level

Keywords

distributed databases; fault tolerant computing; file servers; multiprocessing systems; relational databases; software reliability; HypRa; coarse grained hardware platform; data fragmentation; dynamic reconfigurability; dynamic reconfiguration; dynamic rerouting; extreme requirements; fault detection; fault masking; fault-tolerant multiprocessor database server; fragment replication; multiway interconnection network; replica allocation; Application software; Fault detection; Fault tolerance; Fault tolerant systems; Hardware; Law; Legal factors; Network servers; Telecommunications; Transaction databases;

fLanguage

English

Publisher

ieee

Conference_Titel

Fault Tolerant Systems, 1991. Proceedings., Pacific Rim International Symposium on

Conference_Location

Kawasaki

Print_ISBN

0-8186-2275-X

Type

conf

DOI

10.1109/{RFTS.1991.212941

Filename

212941