Section: New Results

Social Network Analysis on Multi-Core Architectures

Participants : Thomas Messi Nguele, Maurice Tchuente [Univ Yaoundé 1, LIRIMA] , Jean-Francois Mehaut.

One of social graph properties is the community structure, that is, subsets where nodes belonging to the same subset have a higher link density between themselves and a low link density with nodes belonging to external subsets. Furthermore, most social network mining algorithms comprise a local exploration of the underlying graph, which consists in referencing nodes in the neighborhood of a particular node.

The main contribution of this work is to use the community structure during the storage of large graphs that arise in social network mining. The goal is to reduce cache misses and consequently, execution time. After formalizing the problem of social network ordering as a problem of optimal linear arrangement which is known as NP-Complete, we propose NumBaCo, a heuristic based on the community structure. We present, for Katz score and Pagerank, simulations that compare classic data structures Bloc and Yale to their corresponding versions that use NumBaCo. Results on a 32-cores NUMA machine using real datasets (amazon, dblp and web-google) show that NumBaCo allows to reduce from 62% to 80% of cache misses and from 15% to 50% of execution time.

This work was initiated inside the LIRIMA Inria International Laboratory with the University of Yaoundé and Maurice Tchuenté. Those results are part of Thomas Messi Nguélé' PhD which is prepared with a Cotutelle agreement. Those results have been presented at the ParCO international conference [19] and published in the ARIMA (Revue Africaine de Recherche en Informatique et Mathématiques Appliquées) journal [3].