MalleTrain: Deep Neural Networks Training on Unfillable Supercomputer Nodes
Open Access
- 7 May 2024
- conference paper
- Published by Association for Computing Machinery (ACM)
- p. 190-200
Abstract
First-come first-serve scheduling can result in substantial (up to 10%) of transiently idle nodes on supercomputers. Recognizing that such unfilled nodes are well-suited for deep neural network (DNN) training, due to the flexible nature of DNN training tasks, Liu et al. proposed that the re-scaling DNN training tasks to fit gaps in schedules be formulated as a mixed-integer linear programming (MILP) problem, and demonstrated via simulation the potential benefits of the approach. Here, we introduce MalleTrain, a system that provides the first practical implementation of this approach and that furthermore generalizes it by allowing it to be used even for DNN training applications for which model information is unknown before runtime. Key to this latter innovation is the use of a lightweight online job profiling advisor (JPA) to collect critical scalability information for DNN jobs---information that it then employs to optimize resource allocations dynamically, in real time. We describe the MalleTrain architecture and present the results of a detailed experimental evaluation on a supercomputer GPU cluster and several representative DNN training workloads, including neural architecture search and hyperparameter optimization. Our results not only confirm the practical feasibility of leveraging idle supercomputer nodes for DNN training but improve significantly on prior results, improving training throughput by up to 22.3% without requiring users to provide job scalability information.Keywords
All Related Versions
Funding Information
- Department of Energy, Office of Science (DE-AC02-06CH11357)
- NSF (National Science Foundation) (CAREER-2048044)
This publication has 19 references indexed in Scilit:
- FreeTrain: A Framework to Utilize Unused Supercomputer Nodes for Training Neural Networks2023
- Full Waveform Inversion-Based Ultrasound Computed Tomography Acceleration Using Two-Dimensional Convolutional Neural NetworksJournal of Nondestructive Evaluation, Diagnostics and Prognostics of Engineering Systems, 2023
- fairDMS: Rapid Model Training by Data and Model Reuse2022
- Deep Learning Accelerated Light Source Experiments2019
- The Evolution of the Pegasus Workflow Management SoftwareComputing in Science & Engineering, 2019
- Learning Transferable Architectures for Scalable Image Recognition2018
- Parallel Programming with Migratable Objects: Charm++ in Practice2014
- Malleable applications for scalable high performance computingCluster Computing, 2007
- SLURM: Simple Linux Utility for Resource Management2003
- Utilization, predictability, workloads, and user runtime estimates in scheduling the IBM SP2 with backfillingIEEE Transactions on Parallel and Distributed Systems, 2001