Engineering a Scalable High Quality Graph Partitioner

TL;DR

This paper presents a scalable parallel graph partitioning method that significantly improves solution quality on benchmark instances by enhancing various algorithmic components and parallelizing the FM local search.

Contribution

It introduces a scalable parallel framework for high-quality graph partitioning, including novel parallelization of the FM local search algorithm and improved heuristics.

Findings

Achieved better partition quality than previous systems on Walshaw's benchmarks.

Demonstrated scalability to hundreds of processors.

Improved local search heuristics and edge contraction prioritization.

Abstract

We describe an approach to parallel graph partitioning that scales to hundreds of processors and produces a high solution quality. For example, for many instances from Walshaw's benchmark collection we improve the best known partitioning. We use the well known framework of multi-level graph partitioning. All components are implemented by scalable parallel algorithms. Quality improvements compared to previous systems are due to better prioritization of edges to be contracted, better approximation algorithms for identifying matchings, better local search heuristics, and perhaps most notably, a parallelization of the FM local search algorithm that works more locally than previous approaches.