TY - GEN
T1 - Graph Neural Network Training and Data Tiering
AU - Min, Seung Won
AU - Wu, Kun
AU - Hidayetoǧlu, Mert
AU - Xiong, Jinjun
AU - Song, Xiang
AU - Hwu, Wen Mei
N1 - This work was partially supported by the IBM-ILLINOIS Center for Cognitive Computing Systems Research (C3SR). The authors would also like to thank the anonymous reviewers for their constructive feedback.
PY - 2022/8/14
Y1 - 2022/8/14
N2 - Graph Neural Networks (GNNs) have shown success in learning from graph-structured data, with applications to fraud detection, recommendation, and knowledge graph reasoning. However, training GNN efficiently is challenging because: 1) GPU memory capacity is limited and can be insufficient for large datasets, and 2) the graph-based data structure causes irregular data access patterns. In this work, we provide a method to statistically analyze and identify more frequently accessed data ahead of GNN training. Our data tiering method not only utilizes the structure of input graph, but also an insight gained from actual GNN training process to achieve a higher prediction result. With our data tiering method, we additionally provide a new data placement and access strategy to further minimize the CPU-GPU communication overhead. We also take into account of multi-GPU GNN training as well and we demonstrate the effectiveness of our strategy in a multi-GPU system. The evaluation results show that our work reduces CPU-GPU traffic by 87-95% and improves the training speed of GNN over the existing solutions by 1.6-2.1x on graphs with hundreds of millions of nodes and billions of edges.
AB - Graph Neural Networks (GNNs) have shown success in learning from graph-structured data, with applications to fraud detection, recommendation, and knowledge graph reasoning. However, training GNN efficiently is challenging because: 1) GPU memory capacity is limited and can be insufficient for large datasets, and 2) the graph-based data structure causes irregular data access patterns. In this work, we provide a method to statistically analyze and identify more frequently accessed data ahead of GNN training. Our data tiering method not only utilizes the structure of input graph, but also an insight gained from actual GNN training process to achieve a higher prediction result. With our data tiering method, we additionally provide a new data placement and access strategy to further minimize the CPU-GPU communication overhead. We also take into account of multi-GPU GNN training as well and we demonstrate the effectiveness of our strategy in a multi-GPU system. The evaluation results show that our work reduces CPU-GPU traffic by 87-95% and improves the training speed of GNN over the existing solutions by 1.6-2.1x on graphs with hundreds of millions of nodes and billions of edges.
KW - GPU-acceleration
KW - graph neural networks
KW - very large data
UR - http://www.scopus.com/inward/record.url?scp=85137147638&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85137147638&partnerID=8YFLogxK
U2 - 10.1145/3534678.3539038
DO - 10.1145/3534678.3539038
M3 - Conference contribution
AN - SCOPUS:85137147638
T3 - Proceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining
SP - 3555
EP - 3565
BT - KDD 2022 - Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining
PB - Association for Computing Machinery
T2 - 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, KDD 2022
Y2 - 14 August 2022 through 18 August 2022
ER -