A MIGRATION TOOL TO SUPPORT RESOURCE AND LOAD SHARING IN HETEROGENEOUS COMPUTING ENVIRONMENTS

Citation
Pd. Le et B. Srinivasan, A MIGRATION TOOL TO SUPPORT RESOURCE AND LOAD SHARING IN HETEROGENEOUS COMPUTING ENVIRONMENTS, Computer communications, 20(5), 1997, pp. 361-375
Citations number
52
Categorie Soggetti
Computer Sciences","Computer Science Hardware & Architecture","Computer Science Software Graphycs Programming
Journal title
ISSN journal
01403664
Volume
20
Issue
5
Year of publication
1997
Pages
361 - 375
Database
ISI
SICI code
0140-3664(1997)20:5<361:AMTTSR>2.0.ZU;2-#
Abstract
Distributed computing systems potentially provide significant advantag es and benefits such as enhanced performance, extensibility, reliabili ty, and better resource and load sharing. However, the evolution of ha rdware and software has unavoidably]ed to heterogeneous computing syst ems. The heterogeneity of such systems gives rise to many problems suc h as transparency, application developments, database and network mana gement, efficient resource and load sharing, migration, etc. This pape r discusses the problems associated with resource and load sharing in such heterogeneous environments. First, it discusses different techniq ues such as distributed file systems, network operating systems, load- sharing algorithms, migration mechanisms and other methods which have been proposed as solutions for resource and load sharing. Then it desc ribes our design and implementation of a prototype tool for program an d data migration in heterogeneous computing environments of different computer architectures and different operating systems. The tool was b ased on the client-sen;er model and the use of remote procedure calls for interconnecting heterogeneous distributed computer systems. This i mportant decision supports the evolution of the computer hardware and software without side-effects. It adopts well with the evolution of di stributed systems as the degree of heterogeneity increases. The tool a ims to allow better resource and load sharing in heterogeneous compute r systems; to allow useful information to be quickly obtained; to redu ce communication costs and programmers' time in prototyping distribute d applications. It is designed and implemented on a local network of c omputers connected by Ethernet and running different Unix and Unix-lik e operating systems. (C) 1997 Elsevier Science B.V.