MG-GCN: Scalable Multi-GPU GCN Training Framework

TL;DR

MG-GCN is a multi-GPU training framework for Graph Convolutional Networks that overcomes memory and communication bottlenecks, enabling scalable training on large graphs with improved speedup over existing methods.

Contribution

It introduces a multi-GPU GCN training framework with HPC optimizations, allowing scalable training on large graphs and achieving superior speedup compared to state-of-the-art.

Findings

Achieves super-linear speedup on large graphs.

Enables training on datasets that do not fit in a single GPU.

Reduces memory footprint through buffer reuse.

Abstract

Full batch training of Graph Convolutional Network (GCN) models is not feasible on a single GPU for large graphs containing tens of millions of vertices or more. Recent work has shown that, for the graphs used in the machine learning community, communication becomes a bottleneck and scaling is blocked outside of the single machine regime. Thus, we propose MG-GCN, a multi-GPU GCN training framework taking advantage of the high-speed communication links between the GPUs present in multi-GPU systems. MG-GCN employs multiple High-Performance Computing optimizations, including efficient re-use of memory buffers to reduce the memory footprint of training GNN models, as well as communication and computation overlap. These optimizations enable execution on larger datasets, that generally do not fit into memory of a single GPU in state-of-the-art implementations. Furthermore, they contribute to achieve superior speedup compared to the state-of-the-art. For example, MG-GCN achieves super-linear speedup with respect to DGL, on the Reddit graph on both DGX-1 (V100) and DGX-A100.