Distributed-Memory Breadth-First Search on Massive Graphs

TL;DR

This paper analyzes distributed-memory BFS algorithms on supercomputers, comparing traditional and direction-optimizing methods, and evaluates various data structures, threading, and graph decompositions for scalability.

Contribution

It provides a comprehensive performance and scalability analysis of BFS algorithms on large distributed graphs, including new insights into data structures and decomposition strategies.

Findings

Direction-optimizing BFS improves performance on large graphs.

2D graph decomposition enhances scalability over 1D.

CSR and DCSC data structures impact in-node multithreading efficiency.

Abstract

This chapter studies the problem of traversing large graphs using the breadth-first search order on distributed-memory supercomputers. We consider both the traditional level-synchronous top-down algorithm as well as the recently discovered direction optimizing algorithm. We analyze the performance and scalability trade-offs in using different local data structures such as CSR and DCSC, enabling in-node multithreading, and graph decompositions such as 1D and 2D decomposition.