NeutronTP: Load-Balanced Distributed Full-Graph GNN Training with Tensor Parallelism

TL;DR

NeutronTP introduces a tensor parallelism approach for distributed GNN training that balances workload and reduces communication overhead, enabling efficient training of large-scale graphs across multiple GPUs.

Contribution

The paper proposes a novel tensor parallelism method for distributed GNN training that eliminates cross-worker dependencies and improves load balancing and efficiency.

Findings

Achieves up to 8.72x speedup over existing systems

Effectively trains large graphs exceeding single GPU memory

Reduces communication overhead through decoupled training framework

Abstract

Graph neural networks (GNNs) have emerged as a promising direction. Training large-scale graphs that relies on distributed computing power poses new challenges. Existing distributed GNN systems leverage data parallelism by partitioning the input graph and distributing it to multiple workers. However, due to the irregular nature of the graph structure, existing distributed approaches suffer from unbalanced workloads and high overhead in managing cross-worker vertex dependencies. In this paper, we leverage tensor parallelism for distributed GNN training. GNN tensor parallelism eliminates cross-worker vertex dependencies by partitioning features instead of graph structures. Different workers are assigned training tasks on different feature slices with the same dimensional size, leading to a complete load balance. We achieve efficient GNN tensor parallelism through two critical functions. Firstly, we employ a generalized decoupled training framework to decouple NN operations from graph aggregation operations, significantly reducing the communication overhead caused by NN operations which must be computed using complete features. Secondly, we employ a memory-efficient task scheduling strategy to support the training of large graphs exceeding single GPU memory, while further improving performance by overlapping communication and computation. By integrating the above techniques, we propose a distributed GNN training system NeutronTP. Our experimental results on a 16-node Aliyun cluster demonstrate that NeutronTP achieves 1.29X-8.72X speedup over state-of-the-art GNN systems including DistDGL, NeutronStar, and Sancus.