This paper proposes BGL, a distributed GNN training system designed to address the bottlenecks with a few key ideas.

In this paper, we propose BGL, a GPU-efficient GNN train- ing system for large graph learning, to accelerate training and achieve high GPU utilization (near 100 ...

This paper proposes BGL, a distributed GNN training system designed to address the bottlenecks with a few key ideas, including a dynamic cache engine to ...

This is the source code of BGL: GPU-Efficient GNN Training by Optimizing Graph Data I/O and Preprocessing, published in NSDI 2023.

Sep 7, 2024 · The main bottlenecks are the process of preparing data for GPUs - subgraph sampling and feature retrieving. This paper proposes BGL, a ...

Explore a conference talk that delves into BGL, a distributed GNN training system designed to optimize GPU efficiency for large-scale graph data processing.

主要瓶颈是为GPU 准备数据的过程——子图采样和特征检索。本文提出了BGL，这是一种分布式GNN 训练系统，旨在通过一些关键思想解决瓶颈问题。首先，我们提出了 ...

Nov 7, 2024 · BGL:GPU-efficient GNN training by optimizing graph data I/O and preprocessing. In Proceedings of NSDI, 2023. Google Scholar. [5]. Jie Chen ...

GNN Dataloaders. Venue, Title, Affiliation, Link, Source. NSDI 2023, BGL: GPU-Efficient GNN Training by Optimizing Graph Data I/O and Preprocessing, ByteDance ...

Extensive experiments on various GNN models and large graph datasets show that BGL significantly outperforms existing GNN training systems by 20. 68x on average ...