Optimal Scheduling Results for Parallel Computing

Document type: Bookchapters
Peer reviewed: Yes
Author(s): Håkan Lennerstad, Lars Lundberg
Title: Optimal Scheduling Results for Parallel Computing
Translated title: Resultat om optimal scheduling för parallelldatorer
Book: Applications on advanced architecture computers
Year: 1996
Pagination: 155-164
Editor: Greg Astfalk
ISBN: 0-89871-368-4
Publisher: SIAM
City: Philadelphia, USA
Organization: Blekinge Institute of Technology
Department: Dept. of Telecommunications and Mathematics (Institutionen för telekommunikation och matematik)
Dept. of Telecommunications and Mathematics S-37179 Karlskrona
+46 455 38 50 00
Authors e-mail: hln@bth.se, llu@bth.se
Language: English
Abstract: Load balancing is one of many possible causes of poor performance on parallel machines. If good load balancing of the decomposed algorithm or data is not achieved, much of the potential gain of the parallel algorithm is lost to idle processors. Each of the two extremes of load balancing - static allocation and dynamic allocation - has advantages and disadvantages. This chapter illustrates the relationship between static and dynamic allocation of tasks.
Summary in Swedish: Lastbalansering är en av många möjliga orsaker för låga prestanda på parallelldatorer. Var och en av de två extremerna för lastbalansering, statisk allokering och dynamisk allokering, har sina för- och nackdelar. Detta kapitel illustrerar förhållandet mellan dem.
Subject: Computer Science\Computersystems
Mathematics\Discrete Mathematics
Keywords: multiprocessor, load balancing, extremal combinatorics
Edit