Acquisition of a GPU-Accelerated Deep-Learning Research Cluster
Abstract
The intent of this proposal is for the acquisition of a GPU-Accelerated Deep-Learning Research Cluster. The proposed instrumentation is a 24 GPU cluster consisting of two NVIDIA DGX A100 (8x NVIDIA A100 GPUs) research nodes and two 4x NVIDIA A100 GPUs education nodes. The instrumentation has enough computational power to simultaneously run multiple deep-learning training processes and can also efficiently utilize multiple GPUs to train huge models that cannot be trained on a single GPU machine. The instrumentation will also be used to enhance the undergraduate and graduate engineering education experience by providing and enabling virtualized computing services to the students of Virginia State University (VSU), an HBCU minority-serving institution with engineering, science and technology programs. With the proposed instrumentation, VSU aims to invigorate various research projects currently conducted by the members of the Computer Science faculty. Artificial Intelligence (AI) and Machine Learning (ML) research at VSU has been severely hindered by the lack of an instrumentation that provides rapid deep-learning model training and adequate space to store big-data. Consequently, due to resource limitations, faculty oftentimes had to scale down the scope of their projects. Such instrumentation will be essential in scaling up much research conducted at VSU. In addition, the instrumentation will enable VSU faculty to establish new lines of research such as neuromorphic computing simulation. Moreover, the instrumentation will help various ongoing and future collaborative research projects conducted by VSU faculty and partners in academia as well as in industry. The Computer Science department at VSU foresees this proposed instrumentation becoming a cornerstone in establishing the Center for Deep-Learning, one of the departmental objectives in the next five-year plan. When established, the Center will synergistically bring together VSU faculty, their collaborators and their research projects around the centralized computing resources. Additionally, this will help to bridge the gap between fundamental AL, ML, Data Science and Cyber Security research typically performed at universities and application development routinely performed in industries. In particular, the College of Engineering and Technology at VSU has active external research collaborations with member industries and higher education institutions of the Commonwealth Center for Advanced Manufacturing (CCAM). As a member, VSUÕs researchers are engaged in the acceleration and transition of research innovation from laboratory to commercial use. The Center for Deep-Learning will greatly benefit VSU as well as CCAM. Lastly, the GPU-Accelerated Deep-Learning Research Cluster will open up new avenues of undergraduate and graduate research and student education. The proposed instrumentation is not only essential to performing the complex and numerous computations required for the fields of big data and deep learning, but also can provide remote access to virtualized computing environments to the students who cannot otherwise afford to have powerful computing devices to maximize their success in the Computer Science program under the situations like COVID-19. This virtualized computing resources will then be used in the outreach activities that are designed to promote and provide Computer Science education to underrepresented groups of K-12 students to create the pipeline from K-12 to minority serving undergraduate and graduate level programs.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jun 25, 2021
- Source ID
- W911NF2110280
Entities
People
- Wookjin Choi
Organizations
- Army Contracting Command
- Office of the Secretary of Defense
- Virginia State University