• 検索結果がありません。

Japan Advanced Institute of Science and Technology

N/A
N/A
Protected

Academic year: 2021

シェア "Japan Advanced Institute of Science and Technology"

Copied!
3
0
0

読み込み中.... (全文を見る)

全文

(1)

Japan Advanced Institute of Science and Technology

JAIST Repository

https://dspace.jaist.ac.jp/

Title

分散環境におけるフォールトトレラントソフトウエア

の構成法に関する研究

Author(s)

伊関, 浩

Citation

Issue Date

1997‑03

Type

Thesis or Dissertation

Text version

author

URL

http://hdl.handle.net/10119/1014

Rights

Description

Supervisor:片山 卓也, 情報科学研究科, 修士

(2)

in Distributed System

Ko Iseki

Scho ol of Information Science,

Japan Advanced Institute of Science and Technology

February 14,1997

Keywords: fault tolerance,software faulttolerance,FTAG,CML.

The increased useof computersand ourincreased relianceonthem haveledtoaneed

for highly reliable computer systems. Once failure o ccured in computer system, weneed

the system that to guard the failure inuence the whole of system. Even if system has

faluire init ,system isgiving the service. Then this prop ertyis calledfault tolerant. The

goal of fault tolerance isto avoid system failure,evenif faultsare present. There are two

fault tolerance approach: hardware fault tolerance and software fault tolerance. In this

paper, we focus on software fault tolerance. Software fault tolerance is a structure that

constructs to continue providing service despite the existence of software faults and/or

operational faults.

Building fault tolerance software by functional approach, there is many advantage

compared with imperative approach. FTAG(Fault Tolerant AttributeGrammers) model

is a model that functional model for writing fault tolerant software that is based at-

tribute grammers. As described there,this mo del can facilitate the writing of programs

that use various mechanismsfor dealingwith software faultsand/or faults inthe under-

lying computation platform. These include recovery blocks,N{version programing and

process replication. With this approach , a program is written as a series of module do-

compositions,withprovisionsfor redoingandreplicatingmodulesused toimplementfault

tolerance requirements.

This FTAG model ,however, is not yet implemented completely computer systems.

Therfore,itis necessarytoimplementFTAGmodeltocomputer system, and makesome

experience onit.

The goalof this paperistoshowavarageoffaulttolerance basedonfunctionalmo del

andhowtoconstructafaulttolerantsoftwarebyimplementationonlooselycoupledmulti

processor system.

Copyright c

1997byKoIseki

(3)

tions called mo dules. Computing order of modules is determined by attributes' dep en-

dence. And, mo dules that no dependency of attributes is possible that running in con-

currently. Accordingly, it is to be desired that the language to implement FTAG have

architecture of concurrency. In this paper, we will use CML(concurrentML).

CML is a language for concurrent programing. It is an extension of Standard ML

(SML), and is implemented on top of Standerd ML of New Jersey(SML/NJ). A CML

program consists of a number of threads, which use message passing on typ ed channels

tocommunicate. Threadsand channelscan create dynamically,this propretyis useful to

simulate attributes owbetween the threads.

The conversion method forFTAG programinglist isconvert to CMLare as follows.

(1). FTAGprogramisconvertedCMLprogramthathaveamanagementoffaulttolerant

software. This program is actual apply of application software.And this program detect

failure and raise redoing.

(2). Workspacemanager and nodemanager is created thatmanage attributesstored ob-

ject base isrequired by faulttolerant structure redoingand replication.

(3). Composing these twomanagers and CMLprogram whichconvertedbefore, generate

the programwhichrun onCMLinterpreter. Then,thisprogram isfaulttolerantsoftware

in CML.

We call three component that describe above, no de manager, work space manager,

and application applier. These components are important component in fault tolerant

softwareinCML.Eachcomp onentsare communicatewithother componentbychannels.

Channel ismessagepassingwaybetween thread. Inthis system,channelisusedbynodes

of comptational tree to carryattribute values.

No demanagercontrolsthe treethatcreatedbyapplicationappliertoapply functions.

Nodemanagerstoredshapeofcomptationtree. Thenredoingoccuredbydetectingfailure

,make tree to no contradictionin system, redoing operation is execute. Workspaceman-

agerisstore andmanageattribute valuesrequiredwhenredoingand replicationhapp ens.

Application applier is that converted program FTAG to CML. Actually,this component

execute the program of CML.

Messages from the node manager to the workspace manager are required that read

attritute value and require to delete workspace. Messages from work spaces to the node

managerarereportedthatrequiredattributevaluestotheworkspaces. Messagesbetween

other comp oses are required and reported asavob e.

Indecompositionmodules tothreadswithchannels,eachnodecanworkconcurrently.

By this property,this model is possible to implement to distributed system. Based the

architecture we describe above,we show this implementation FTAG model is applicable

to destributed system. And through description of name sarver system, in conrm the

eciency of FTAG model, and itsimplementation of ondistributed systems.

参照

関連したドキュメント

Precisely, over a period of 120 months, the total number of new infections that will be generated from the two patches in the absence of optimal control is 1.2037× 10 4 , whereas,

It is known that minimal Sullivan models for a simply connected space of finite type are all isomorphic, and that the isomorphism class of a minimal Sullivan model for a

We construct a cofibrantly generated model structure on the category of flows such that any flow is fibrant and such that two cofibrant flows are homotopy equivalent for this

We define the notion of an additive model category and prove that any stable, additive, combinatorial model category M has a model enrichment over Sp Σ (s A b) (symmetric spectra

In this paper, for the first time an economic production quantity model for deteriorating items has been considered under inflation and time discounting over a stochastic time

3.1, together with the result in (Barber and Plotkin 1997) (completeness via the term model construction), is that the term model of DCLL forms a model of DILL, i.e., a

Hence for a given multiscale network, we may perform many steps of model reduction until we obtain a reduced model which is simple enough to allow for extensive simulations, that

At the same time, a new multiplicative noise removal algorithm based on fourth-order PDE model is proposed for the restoration of noisy image.. To apply the proposed model for