Comprehensive Evaluation of GNN Training Systems: A Data Management Perspective

arXiv:2311.13279v2 Announce Type: replace
Abstract: Many Graph Neural Network (GNN) training systems have emerged recently to support efficient GNN training. Since GNNs embody complex data dependencies between training samples, the training of GNNs should address distinct challenges different from DNN training in data management, such as data partitioning, batch preparation for mini-batch training, and data transferring between CPUs and GPUs. These factors, which take up a large proportion of training time, make data management in GNN training more significant. This …

abstract arxiv challenges cs.dc cs.lg data data management data partitioning dependencies dnn evaluation gnn gnns graph graph neural network management network neural network partitioning perspective samples support systems training type