ResearchTrend.AI
  • Papers
  • Communities
  • Events
  • Blog
  • Pricing
Papers
Communities
Social Events
Terms and Conditions
Pricing
Parameter LabParameter LabTwitterGitHubLinkedInBlueskyYoutube

© 2025 ResearchTrend.AI, All rights reserved.

  1. Home
  2. Papers
  3. 2104.06700
25
119

DistGNN: Scalable Distributed Training for Large-Scale Graph Neural Networks

14 April 2021
Vasimuddin
Sanchit Misra
Guixiang Ma
Ramanarayan Mohanty
E. Georganas
A. Heinecke
Dhiraj D. Kalamkar
Nesreen Ahmed
Sasikanth Avancha
    GNN
ArXivPDFHTML
Abstract

Full-batch training on Graph Neural Networks (GNN) to learn the structure of large graphs is a critical problem that needs to scale to hundreds of compute nodes to be feasible. It is challenging due to large memory capacity and bandwidth requirements on a single compute node and high communication volumes across multiple nodes. In this paper, we present DistGNN that optimizes the well-known Deep Graph Library (DGL) for full-batch training on CPU clusters via an efficient shared memory implementation, communication reduction using a minimum vertex-cut graph partitioning algorithm and communication avoidance using a family of delayed-update algorithms. Our results on four common GNN benchmark datasets: Reddit, OGB-Products, OGB-Papers and Proteins, show up to 3.7x speed-up using a single CPU socket and up to 97x speed-up using 128 CPU sockets, respectively, over baseline DGL implementations running on a single CPU socket

View on arXiv
Comments on this paper