Comprehensive Evaluation of GNN Training Systems: A Data Management Perspective

TL;DR

This paper reviews and evaluates GNN training systems focusing on data management challenges, analyzing approaches through extensive experiments and offering practical insights for future system design.

Contribution

It provides a comprehensive analysis of GNN training data management, including experimental evaluation and practical recommendations for system development.

Findings

Data partitioning and transfer significantly impact training efficiency.

Different approaches vary in handling data dependencies and transfer overhead.

Practical tips improve GNN training system performance.

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 paper reviews GNN training from a data management perspective and provides a comprehensive analysis and evaluation of the representative approaches. We conduct extensive experiments on various benchmark datasets and show many interesting and valuable results. We also provide some practical tips learned from these experiments, which are helpful for designing GNN training systems in the future.