SC21 Proceedings

The International Conference for High Performance Computing, Networking, Storage, and Analysis

Efficient Scaling of Dynamic Graph Neural Networks


Authors: Venkatesan T. Chakaravarthy, Shivmaran S. Pandian, Saurabh Raje, and Yogish Sabharwal (IBM Research) and Toyotaro Suzumura and Shashanka Ubaru (IBM TJ Watson Research Center)

Abstract: We present distributed algorithms for training dynamic Graph Neural Networks (GNN) on large scale graphs spanning multi-node, multi-GPU systems. To the best of our knowledge, this is the first scaling study on dynamic GNN. We devise mechanisms for reducing the GPU memory usage and identify two execution time bottlenecks: CPU-GPU data transfer; and communication volume. Exploiting properties of dynamic graphs, we design a graph difference-based strategy to significantly reduce the transfer time. We develop a simple, but effective data distribution technique under which the communication volume remains fixed and linear in the input size, for any number of GPUs. Our experiments using billion-size graphs on a system of 128 GPUs shows that: (i) the distribution scheme achieves up to 30x speedup on 128 GPUs; (ii) the graph-difference technique reduces the transfer time by a factor of up to 4.1x and the overall execution time by up to 40%.


Presentation: file


Back to Technical Papers Archive Listing