Towards Data Mining on Emerging Architectures

Recent advances in microprocessor design have given rise to new commodity architectures. One such innovation is to place multiple cores on a single chip, called Chip Multiprocessing (CMP). Each core is an independent computational unit, allowing multiple processes to execute concurrently. A second recent architectural advancement is to allow multiple processes to compete for resources simultaneously on a single core, called simultaneous multithreading (SMT). SMT can improve overall throughput in cases where CPU utilization is low. We investigate the implications of these advances on the design of data mining algorithms. In particular, we focus on frequent graph mining. Mining graph based data sets has practical applications in many areas including molecular substructure discovery, web link analysis, fraud detection, and social network analysis. In this work, we propose a novel approach for parallelizing graph mining on CMP architectures. We design a parallel algorithm with low memory consumption, low bandwidth, and fine task granularity. We show that dynamic partitioning and dynamic task allocation provide a synergy which greatly improves scalability over a naive algorithm, from 5 fold to 27 fold on 32 nodes.