• 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!
2
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

疎結合分散環境における耐故障性と適応性を実現する

ソフトウェアの構成に関する研究

Author(s)

豊島, 真澄

Citation

Issue Date

2001‑06

Type

Thesis or Dissertation

Text version

author

URL

http://hdl.handle.net/10119/927

Rights

Description

Supervisor:片山 卓也, 情報科学研究科, 博士

(2)

on Loosely Coupled Distributed Systems

Masumi Toyoshima

Schoolof Information Science,

Japan Advanced Institute of Science and Technology

June 2001

Abstract

Thisthesis presentsatask allocationmethod to implementfault tolerant software using functionalpro-

grammingparadigm. Thedesignand implementation usinggroupcommunicationsystemisalso discussed

in ordertobuildtheruntimesystemonlooselycoupleddistributed environment.

Recently,manyCOTScomputersareconnectedusingLANorWANandthereariseseveralprojectswhich

usethecomputationpoweroftheseinexpensivecomputationresources.Intheselooselycoupleddistributed

environment,thereexistsmanycomponentsbuildupsinglesystem,there aredierentclassofperformance

of computers, and the communication links may be sometimes down. In order to run large applications

including longrunningscientic computation onsuch environment,it is important forsuch systemstobe

fault tolerance,thecharacteristicsofasystemtotoleratesomefaultand continuerunninginanacceptable

level.

Manyresearchworkhasbeendoneinthis areasince1970sandthere existssomebasictechniquestoim-

plementfaulttolerance. However,mostofthesetechniquesarebasedonimperativeprogrammingparadigm.

Thoughitiseasytounderstanditsoperationalsemanticsusingimperativeprogrammingparadigm,program-

mers have to consider many complication: detecting the fault, checkpointingsystem state and recovering

usingthem tocorrectstates,etc.

InordertoavoidthesediÆculties,APRreplicationtechniquewhich isbasedonfunctional programming

paradigmisintroducedin1998. Theholedevelopment,fromthemodelofcomputationthroughimplemen-

tationis introducedin APRapproach. APRprovidesnotonlyfaulttolerancebut italsoshortensthetime

to completecomputation. Itis onlyneededforprogrammerstodescribeapplication programin functional

mannerin ordertogaintheabovebenets.

Thoughtheschedulingalgorithm offunctions in APR isintroduced, theresourceallocationmethodnor

details ofdesign and implementation includingcommunicationin looselycoupled distributed environment

arestill notdened.

This theses starts with the formalization of APR task scheduling algorithm. Then the RAFT resource

management system is introduced in order to manage computation resources and allocate tasks to the

resources. RAFT divides APR functions to more ne-grainedtasks called RAFT process and distributes

these processesto thecomputation resourcesexisting on network. Therecovery process in RAFT is also

dened soastominimizetherecoverytimewhenfailuresoccur.

Thedesignandimplementationusinggroupcommunicationisalsointroducedinthisthesiswithitscost

analysis. Theseworksshowtheeectivenessandcharacteristicsofthefaulttolerantsystemwhichbasedon

functional programmingparadigm.

keywords: faulttolerantsoftware,looselycoupleddistributedsystems,parallelcomputation,functional

programming,resourceallocation

参照

関連したドキュメント

“We’d like not just text or diagram, but both!”.

○ only symmetric operations (invariant over permutation of bases/coordinates). Targeted abduction:

[11] Sugiyama S., On some problems on functional differential equations with advanced arguments, Proceedings US-Japan Seminar on Differential and Functional Equations,

In [6] we considered some nonlinear elliptic functional differential equations where we proved theorems on the number of weak solutions of boundary value problems for such equations

What relates to Offline Turing Machines in the same way that functional programming languages relate to Turing Machines?.. Int Construction.. Understand the transition from

This research was supported by Natural Science Foundation of the Higher Education Institutions of Jiangsu Province (10KJB110003) and Jiangsu Uni- versity of Science and

In the proofs we follow the technique developed by Mitidieri and Pohozaev in [6, 7], which allows to prove the nonexistence of not necessarily positive solutions avoiding the use of

This review is devoted to the optimal with respect to accuracy algorithms of the calculation of singular integrals with fixed singu- larity, Cauchy and Hilbert kernels, polysingular