2017 IEEE International Conference on Cluster Computing (CLUSTER) (2017)
Honolulu, Hawaii, United States
Sept. 5, 2017 to Sept. 8, 2017
Graph models of social information systems typically contain trillions of edges. Such big graphs cannot beprocessed on a single machine. The graph object must bepartitioned and distributed among machines and processedin parallel on a computer cluster. Programming such systemsis very challenging. In this work, we present DH-Falcon, a graph DSL (domain-specific language) which can be usedto implement parallel algorithms for large-scale graphs, tar-geting Distributed Heterogeneous (CPU and GPU) clusters. DH-Falcon compiler is built on top of the Falcon compiler, which targets single node devices with CPU and multipleGPUs. An important facility provided by DH-Falcon is that itsupports mutation of graph objects, which allows programmerto write dynamic graph algorithms. Experimental evaluationshows that DH-Falcon matches or outperforms state-of-the-art frameworks and gains a speedup of up to 13×.
Graphics processing units, Heuristic algorithms, Mirrors, DSL, Computational modeling, Partitioning algorithms, Clustering algorithms
U. Cheramangalath, R. Nasre and Y. N. Srikant, "DH-Falcon: A Language for Large-Scale Graph Processing on Distributed Heterogeneous Systems," 2017 IEEE International Conference on Cluster Computing (CLUSTER), Honolulu, Hawaii, United States, 2017, pp. 439-450.