Betweenness centrality on Multi-GPU systems

Bernaschi, Massimo; Carbone, Giancarlo; Vella, Flavio

doi:10.1145/2833179.2833192

Cited by 3 publications

(5 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To the best of our knowledge, there are no solutions, based on a linear algebra formulation, which exploit heuristics to speed-up the BC computation. Bernaschi et al,in [7] proposed the first fully distributed BC on Multi-GPU systems. Their solution scales well up to 64 GPUs on Friendster graph [26].…”

Section: Related Workmentioning

confidence: 99%

“…We propose a solution to overcome that limitation in Section 3.4.1. Moreover, in [7], a distributed 1-degree reduction preprocessing based on 1-D partitioning is described and evaluated.…”

Section: Heuristics For Betweenness Centralitymentioning

confidence: 99%

“…However, in those solutions, the size of the graph is limited by the space complexity of Brandes' algorithm. Pioneering approaches [11,7] overcome the memory limitations by distributing the graph among more computational nodes. In particular, Bernaschi et al, proposed the first fully distributed BC on clusters of Graphics Processing Units (GPU).…”

Section: Introductionmentioning

confidence: 99%

“…Although advanced techniques overcome many of the difficulties (e.g., the use of a bit-mask improves the scalability [41]), a distributed BC implementation requires further/specific techniques due to Brandes' algorithm timecomplexity and for the different memory requirements with respect to a simple BFS. For instance, a distributed BC implementation requires information exchange about the shortest path so there is no benefit from using a bit-mask during the communication [7].…”

Section: Introductionmentioning

confidence: 99%

“…The time reported is obtained on a single node on Piz Daint 7. The mean time is computed considering only connected vertices.…”

mentioning

confidence: 99%

See 4 more Smart Citations

Algorithms and Heuristics for Scalable Betweenness Centrality Computation on Multi-GPU Systems

Vella,

Carbone,

Bernaschi

2016

Preprint

Self Cite

View full text Add to dashboard Cite

Betweenness Centrality (BC) is steadily growing in popularity as a metrics of the influence of a vertex in a graph. The BC score of a vertex is proportional to the number of all-pairs-shortestpaths passing through it. However, complete and exact BC computation for a large-scale graph is an extraordinary challenge that requires high performance computing techniques to provide results in a reasonable amount of time. Our approach combines bi-dimensional (2-D) decomposition of the graph and multi-level parallelism together with a suitable data-thread mapping that overcomes most of the difficulties caused by the irregularity of the computation on GPUs. Furthermore, we propose novel heuristics which exploit the topology information of the graph in order to reduce time and space requirements of BC computation. Experimental results on synthetic and real-world graphs show that the proposed techniques allow the BC computation of graphs which are too large to fit in the memory of a single computational node along with a significant reduction of the computing time.

show abstract

Section: Related Workmentioning

confidence: 99%

“…We propose a solution to overcome that limitation in Section 3.4.1. Moreover, in [7], a distributed 1-degree reduction preprocessing based on 1-D partitioning is described and evaluated.…”

Section: Heuristics For Betweenness Centralitymentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

“…The time reported is obtained on a single node on Piz Daint 7. The mean time is computed considering only connected vertices.…”

mentioning

confidence: 99%

See 3 more Smart Citations