Parallel process placement
This thesis investigates methods of automatic allocation of processes to available processors in a given network configuration. The research described covers the investigation of various algorithms for optimal process allocation. Among those researched were an algorithm which used a branch and bound...
Main Author: | |
---|---|
Format: | Others |
Language: | English |
Published: |
Rhodes University
1989
|
Subjects: | |
Online Access: | http://hdl.handle.net/10962/d1002033 |
id |
ndltd-netd.ac.za-oai-union.ndltd.org-rhodes-vital-4568 |
---|---|
record_format |
oai_dc |
spelling |
ndltd-netd.ac.za-oai-union.ndltd.org-rhodes-vital-45682017-07-20T04:13:14ZParallel process placementHandler, CarolineParallel programming (Computer science)Parallel programming (Computer science)This thesis investigates methods of automatic allocation of processes to available processors in a given network configuration. The research described covers the investigation of various algorithms for optimal process allocation. Among those researched were an algorithm which used a branch and bound technique, an algorithm based on graph theory, and an heuristic algorithm involving cluster analysis. These have been implemented and tested in conjunction with the gathering of performance statistics during program execution, for use in improving subsequent allocations. The system has been implemented on a network of loosely-coupled microcomputers using multi-port serial communication links to simulate a transputer network. The concurrent programming language occam has been implemented, replacing the explicit process allocation constructs with an automatic placement algorithm. This enables the source code to be completely separated from hardware considerationsRhodes UniversityFaculty of Science, Computer Science1989ThesisMastersMSc54 leavespdfvital:4568http://hdl.handle.net/10962/d1002033EnglishHandler, Caroline |
collection |
NDLTD |
language |
English |
format |
Others
|
sources |
NDLTD |
topic |
Parallel programming (Computer science) Parallel programming (Computer science) |
spellingShingle |
Parallel programming (Computer science) Parallel programming (Computer science) Handler, Caroline Parallel process placement |
description |
This thesis investigates methods of automatic allocation of processes to available processors in a given network configuration. The research described covers the investigation of various algorithms for optimal process allocation. Among those researched were an algorithm which used a branch and bound technique, an algorithm based on graph theory, and an heuristic algorithm involving cluster analysis. These have been implemented and tested in conjunction with the gathering of performance statistics during program execution, for use in improving subsequent allocations. The system has been implemented on a network of loosely-coupled microcomputers using multi-port serial communication links to simulate a transputer network. The concurrent programming language occam has been implemented, replacing the explicit process allocation constructs with an automatic placement algorithm. This enables the source code to be completely separated from hardware considerations |
author |
Handler, Caroline |
author_facet |
Handler, Caroline |
author_sort |
Handler, Caroline |
title |
Parallel process placement |
title_short |
Parallel process placement |
title_full |
Parallel process placement |
title_fullStr |
Parallel process placement |
title_full_unstemmed |
Parallel process placement |
title_sort |
parallel process placement |
publisher |
Rhodes University |
publishDate |
1989 |
url |
http://hdl.handle.net/10962/d1002033 |
work_keys_str_mv |
AT handlercaroline parallelprocessplacement |
_version_ |
1718500792815583232 |