A k-way Greedy Graph Partitioning with Initial Fixed Vertices for Parallel Applications

Maria Predari (1, 2, 3) , Aurélien Esnard (3, 1, 2)

1 HiePACS - High-End Parallel Algorithms for Challenging Numerical Simulations
2 LaBRI - Laboratoire Bordelais de Recherche en Informatique
3 UB - Université de Bordeaux

Graph partitioning is used to solve the problem of distributing computations to a number of processors, in order to improve the performance of time consuming applications in parallel environments. A common approach to solve this problem is based on a multilevel framework, where the graph is firstly coarsened to a smaller instance and then it is partitioned in a number of parts using recursive bisection (RB) based methods. However, in applications where initial fixed vertices are used to model additional constraints of the problem, RB based methods often fail to produce partitions of good quality. In this paper, we propose a new direct k-way greedy graph growing algorithm, called KGGGP, that overcomes this issue and succeeds to produce partition with better quality than RB while respecting the constraint of fixed vertices. In the experimental section, we present results which compare KGGGP against state-of-the-art methods for graphs available from the popular DIMACS'10 collection.

Mots clés: multilevel framework; graph partitioning; high-performance computing; load balancing; parallel simulations

Domaines: Calcul parallèle, distribué et partagé [cs.DC]

Maria Predari, Aurélien Esnard. A k-way Greedy Graph Partitioning with Initial Fixed Vertices for Parallel Applications. 24th Euromicro International Conference on Parallel, Distributed, and Network-Based Processing, Feb 2016, Heraklion, Greece. pp.8. ⟨hal-01277392⟩