# Accelerating Sampling and Aggregation Operations in GNN Frameworks with GPU Initiated Direct Storage Accesses

arXiv: https://arxiv.org/abs/2306.16384 · HTML: https://mlsanity.com/paper/2306.16384/

## Key facts

- Cited by 6 papers in the map (data through 14 Oct 2026).
- Builds on 27 of its 54 references that are in the map.
- First posted on arXiv 28 Jun 2023.

## Builds on (most cited first, 27 of 27)

- [Semi-Supervised Classification with Graph Convolutional Networks](https://mlsanity.com/paper/1609.02907/) (2016)
- [Graph Attention Networks](https://mlsanity.com/paper/1710.10903/) (2017)
- [Inductive Representation Learning on Large Graphs](https://mlsanity.com/paper/1706.02216/) (2017)
- [Fast Graph Representation Learning with PyTorch Geometric](https://mlsanity.com/paper/1903.02428/) (2019)
- [Open Graph Benchmark: Datasets for Machine Learning on Graphs](https://mlsanity.com/paper/2005.00687/) (2020)
- [On Large-Batch Training for Deep Learning: Generalization Gap and Sharp Minima](https://mlsanity.com/paper/1609.04836/) (2016)
- [Spectral Networks and Locally Connected Networks on Graphs](https://mlsanity.com/paper/1312.6203/) (2013)
- [Link Prediction Based on Graph Neural Networks](https://mlsanity.com/paper/1802.09691/) (2018)
- [Graph Neural Networks for Social Recommendation](https://mlsanity.com/paper/1902.07243/) (2019)
- [Cluster-GCN: An Efficient Algorithm for Training Deep and Large Graph Convolutional Networks](https://mlsanity.com/paper/1905.07953/) (2019)
- [Few-Shot Learning with Graph Neural Networks](https://mlsanity.com/paper/1711.04043/) (2017)
- [ZeRO-Infinity: Breaking the GPU Memory Wall for Extreme Scale Deep Learning](https://mlsanity.com/paper/2104.07857/) (2021)
- [Layer-Dependent Importance Sampling for Training Deep and Large Graph Convolutional Networks](https://mlsanity.com/paper/1911.07323/) (2019)
- [AliGraph: A Comprehensive Graph Neural Network Platform](https://mlsanity.com/paper/1902.08730/) (2019)
- [DistDGL: Distributed Graph Neural Network Training for Billion-Scale Graphs](https://mlsanity.com/paper/2010.05337/) (2020)
- [Alleviating the Inconsistency Problem of Applying Graph Neural Network to Fraud Detection](https://mlsanity.com/paper/2005.00625/) (2020)
- [PinnerSage: Multi-Modal User Embedding Framework for Recommendations at Pinterest](https://mlsanity.com/paper/2007.03634/) (2020)
- [FeatGraph: A Flexible and Efficient Backend for Graph Neural Network Systems](https://mlsanity.com/paper/2008.11359/) (2020)
- [IGB: Addressing The Gaps In Labeling, Features, Heterogeneity, and Size of Public Graph Datasets for Deep Learning Research](https://mlsanity.com/paper/2302.13522/) (2023)
- [Ginex: SSD-enabled Billion-scale Graph Neural Network Training on a Single Machine via Provably Optimal In-memory Caching](https://mlsanity.com/paper/2208.09151/) (2022)
- [PyTorch-Direct: Enabling GPU Centric Data Access for Very Large Graph Neural Network Training with Irregular Accesses](https://mlsanity.com/paper/2101.07956/) (2021)
- [GPU-Initiated On-Demand High-Throughput Storage Access in the BaM System Architecture](https://mlsanity.com/paper/2203.04910/) (2022)
- [Medical Entity Disambiguation Using Graph Neural Networks](https://mlsanity.com/paper/2104.01488/) (2021)
- [Graph Neural Network Training with Data Tiering](https://mlsanity.com/paper/2111.05894/) (2021)
- [Accurate, Efficient and Scalable Training of Graph Neural Networks](https://mlsanity.com/paper/2010.03166/) (2020)
- [PiPAD: Pipelined and Parallel Dynamic GNN Training on GPUs](https://mlsanity.com/paper/2301.00391/) (2023)
- [ZIPPER: Exploiting Tile- and Operator-level Parallelism for General and Scalable Graph Neural Network Acceleration](https://mlsanity.com/paper/2107.08709/) (2021)

## Built on by (most cited first, 6 of 6)

- [DiskGNN: Bridging I/O Efficiency and Model Accuracy for Out-of-Core GNN Training](https://mlsanity.com/paper/2405.05231/) (2024)
- [LSM-GNN: Large-scale Storage-based Multi-GPU GNN Training by Optimizing Data Transfer Scheme](https://mlsanity.com/paper/2407.15264/) (2024)
- [Accelerating Storage-Based Training for Graph Neural Networks](https://mlsanity.com/paper/2601.01473/) (2026)
- [EmbedPart: Embedding-Driven Graph Partitioning for Scalable Graph Neural Network Training](https://mlsanity.com/paper/2604.01000/) (2026)
- [FeLoG: Scalable and Efficient Distributed Graph Embedding with Feedback Loop Mechanism](https://mlsanity.com/paper/2606.22180/) (2026)
- [Taurus: Accelerating Out-of-Core Graph Neural Network Inference on Billion-Scale Graphs](https://mlsanity.com/paper/2607.17374/) (2026)

## Abstract (arXiv)

Graph Neural Networks (GNNs) are emerging as a powerful tool for learning from graph-structured data and performing sophisticated inference tasks in various application domains. Although GNNs have been shown to be effective on modest-sized graphs, training them on large-scale graphs remains a significant challenge due to lack of efficient data access and data movement methods. Existing frameworks for training GNNs use CPUs for graph sampling and feature aggregation, while the training and updating of model weights are executed on GPUs. However, our in-depth profiling shows the CPUs cannot achieve the throughput required to saturate GNN model training throughput, causing gross under-utilization of expensive GPU resources. Furthermore, when the graph and its embeddings do not fit in the CPU memory, the overhead introduced by the operating system, say for handling page-faults, comes in the critical path of execution. To address these issues, we propose the GPU Initiated Direct Storage Access (GIDS) dataloader, to enable GPU-oriented GNN training for large-scale graphs while efficiently utilizing all hardware resources, such as CPU memory, storage, and GPU memory with a hybrid data placement strategy. By enabling GPU threads to fetch feature vectors directly from storage, GIDS dataloader solves the memory capacity problem for GPU-oriented GNN training. Moreover, GIDS dataloader leverages GPU parallelism to tolerate storage latency and eliminates expensive page-fault overhead. Doing so enables us to design novel optimizations for exploiting locality and increasing effective bandwidth for GNN training. Our evaluation using a single GPU on terabyte-scale GNN datasets shows that GIDS dataloader accelerates the overall DGL GNN training pipeline by up to 392X when compared to the current, state-of-the-art DGL dataloader.