Paper Info
Title
Mesh Partitioning for Computational Grids
Abstract
Computational grids, with a collection of heterogeneousresources connected by heterogeneous networks, are beingdeveloped globally. Efficient partitioning of applications toexploit such grids is becoming an emerging area of research.Recent work in our group led to the creation of thePaGrid, an application that partitions meshes onto grids.The first version of PaGrid used a multilevel graph partitioningapproach, with refinement based on minimizationof total communication cost, augmented by execution timeload balancing in the final uncoarsening phase. PaGrid wasfound to produce significantly better partitions for two modelsof heterogeneous computational grids, compared to twoother partitioners, JOSTLE and METIS. In the present workwe have redesigned PaGrid to minimize the estimated applicationexecution time in all re.nement levels of the multilevelscheme. This has led to signi.cant improvements inpartition quality.
Year
DOI
Venue
2004
10.1109/DNSR.2004.1344745
CNSR
Keywords
Field
DocType
better partition,partitions mesh,cant improvement,estimated applicationexecution time,efficient partitioning,mesh partitioning,execution timeload balancing,heterogeneous network,pagrid wasfound,modelsof heterogeneous computational grid,computational grid,computational grids,minimisation,grid computing,concurrent computing,graph partitioning,computer networks,load balance,cost function,heterogeneous computing,application software,heterogeneous networks,computer science,mesh generation,finite element methods,computer applications,computational modeling,minimization
Load management,Polygon mesh,Grid computing,Computer science,Load balancing (computing),Parallel computing,Concurrent computing,Heterogeneous network,Graph partition,Mesh generation,Distributed computing
Conference
ISBN
Citations 
PageRank 
0-7695-2096-0
2
0.41
References 
Authors
6
2
Name
Order
Citations
PageRank
Renaud Wanschoor170.88
Eric Aubanel2579.75