The CYBER cluster provides the essential cyberinfrastructure for operations, storage, computation, access, and structuring of data. More importantly, it forms the basis for the exciting diffusion of artificial intelligence, machine learning, and data science analysis into HS through the design of the cyberinfrastructure (storage, memory, compute clusters for ML/DS applications) and the capacity of experts, who will execute projects and train investigators in those methods, enabling rapid diffusion of ideas and techniques.
CYBER
EduceLab Compute Cluster (ECC)
The EduceLab Compute Cluster is an HPC cluster with 5 GPU compute nodes providing 20 NVIDIA H100 GPUs (80GB VRAM), 640 cores, and 4.25TB RAM, backed by a Lenovo GPFS (DSS-G) parallel file system, login nodes, a DTN node, and admin nodes. It offers GPU-accelerated compute for scientific analysis of collected data using AI/ML and HPC processing software.
|
Node Type |
Nodes |
Processor |
Cores/node |
Memory/node |
GPU |
Network |
|
Admin (Mothership) |
2 |
Intel Xeon Platinum 8358 (Ice Lake-SP) @ 2.60GHz |
64 |
256 GB |
— |
InfiniBand HDR (100Gbps) |
|
Login |
2 |
Intel Xeon Platinum 8358 (Ice Lake-SP) @ 2.60GHz |
64 |
512 GB |
— |
InfiniBand HDR (100Gbps) |
|
Compute |
5 |
Intel Xeon Platinum 8358 (Ice Lake-SP) @ 2.60GHz |
64 |
512 GB |
4× NVIDIA H100, Gen 5 PCIe, 80GB (20 total) |
InfiniBand HDR (100Gbps) |
|
Data Transfer (DTN) |
1 |
Intel Xeon Platinum 8358 (Ice Lake-SP) @ 2.60GHz |
64 |
256 GB |
— |
10Gbps Ethernet (external) / HDR (100Gbps internal) |
Storage: Lenovo GPFS (DSS-G) parallel file system — 1.64PB raw, ~1.28PB usable
EduceLab Data Storage
EduceLab Data Storage includes a 4PB NAS RAID disk storage system and file/Globus server nodes. It provides large-scale, reliable disk storage for collected and processed data which can be shared securely through Globus.
High-Speed Network
The high-speed network supports rapid data transfer between ECC compute nodes and EduceLab cluster instruments for imaging, analysis, and data-intensive research. The network is designed to move large datasets efficiently between instruments, storage, and computational resources, reducing delays between capture, processing, and interpretation. The system includes a 100 Gbps InfiniBand cluster interconnect, a 100 Gbps single-mode optical connection to the instrument cluster network switch, and a 100 Gbps management and internet interconnect switch.
System Services Cluster
The System Services Cluster contains a 3-node ProxMox virtualization cluster with an NVIDIA RTX PRO 6000 GPU offering 96GB of VRAM, 64 cores, 512GB RAM, and 2TB of disk. Its uses span user login services, web pages and content, and web portals with advanced UI for publishing heritage science research and complex web schemes.