English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

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

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