Institute for Data Engineering and Science (IDEaS)
2023-24 Call for Profosals for Cyberinfrastructure Resources
The Institute for Data Engineering and Science (IDEaS) is soliciting proposals from individual academic/research faculty or groups engaged in joint projects/centers to seek cyberinfrastructure resource support for their research needs. Over the years, IDEaS invested in a small set of core cyberinfrastructure resources to support its research mission. The resources are managed by the PACE group (https://pace.gatech.edu/). Available resources are:
- The Hive high-performance cluster: A 484-node cluster with dual 12-core Xeon 6226 CPUs and 192 GB/node RAM connected by 100 Gb/sec EDR Infiniband. The cluster contains several nodes with augmented capabilities such as 16 with large SSDs, 16 with large local disks, 16 with quad Nvidia V100 GPUs, and 4 with high RAM ranging from 1.5-3TB (https://docs.pace.gatech.edu/hive/resources/). Ideal for traditional HPC workloads.
- AMD Genoa cluster: A 4-node high-memory cluster with dual 64-core AMD Epyc 9534 CPUs, dual InfiniBand NICs, and 3 TB/node RAM for a 512-core shared memory system with 12 TB total memory. Ideal for big data analysis involving irregular memory accesses.
- NVIDIA DGX H-100: 8 NVIDIA H100 Tensor Core GPUs and 640 GB HBM3 memory supported by dual 56-core Xeon CPUs and 2 TB RAM. Ideal for ML/AI workloads such as NN training, Tensor Libraries, and LLMs.
- petascale storage repository: Expandable DDN storage repository with 1.25 PB current capacity. Ideal for long-term storage of big data sets.
Proposers can request access to a combination of these resources as needed for their research. IDEaS also provides access to cloud resources through a partnership with Microsoft, which were distributed in a separate call earlier. In keeping with IDEaS FY24 thematic focus on AI, projects that target Artificial Intelligence (AI) and its applications receive higher priority, including but not limited to: Foundational topics in AI; Development of software for AI; AI-driven research in sciences, engineering, finance, humanities, arts, and sustainability; and development or storage of datasets to enable such research.
To apply, please submit a pdf document of at most 2 pages that clearly articulates the research needs, anticipated impact of the proposed research, and a detailed listing of which resources are needed and the corresponding estimates of extent of usage (e.g., CPU hours, GPU hours, storage capacity, etc.). The proposals should be submitted by 5pm on September 22nd, to email@example.com with the topic of the proposal in the subject line.