Big Data Science Center

Scientific Scope
The BDSC at the SSRF deployed the first Superfacility in China, and one of the first worldwide. With its Artificial Intelligence-SSRF-Superfacility Platform (AI-SSRF-SP), the BDSC has implemented a centralized Big Data framework capable of tightly integrating multidisciplinary experiments at the large scientific facilities with high-performance computing (HPC), algorithm implementation, scientific theories, Big Data InfiniBand network infrastructures, remote access, beamline control and real-time data pipeline systems, auxiliary laboratory systems and user management systems. Here, to increase the rate of the users’ scientific discoveries and the resulting technological advancements, petabytes of scientific data and metadata from all the experiments at SSRF are, in real-time, collected, processed, analyzed, interpretated and visualized, as well as tagged and tracked by the AI-SSRF-SP to be then ingested by neural networks for Machine Learning (ML), in order to fully automate the entire large facility lifecycle. The BDSC, thus, allows all the users to fully exploit the scientific and technological potential of the SSRF for their research, shifting their focus from data to science, and enabling real-time unmanned multimodal experiments, where massive support in terms of scientific computing is required to enable the most complete knowledge transfer from scientific research to industrial developments for real world applications, while elastically scaling out the workloads on the top Chinese National Supercomputers.
 
 
HPC Infrastructure
The BDSC has designed and deployed the state-of-the-art Big Data science and technologies, hosting AI, Internet of Things (IoT), real-time unstaffed and remotely controlled experiments, robotic automation, HPC, cloud / edge supercomputing, a unified and synchronized storage system, and heterogeneous CPU / GPU / FAT massive parallelisation, within a fully centralized and user-centric platform, which accelerated the users’ multidisciplinary science carried out at the SSRF by several orders of magnitude.
 
 
Specifications
 1 PFlop/s  11000 Cores
 23 PB Storage  53.4 TB Memory
 CPU
208 Intel Xeon Platinum & Gold Nodes
4 Dual-CPU PCIe Nodes
 InfiniBand Network
 GPU
9 Nodes
1 Quad-GPU NVLINK Node
12 NVIDIA Tesla P100
16 NVIDIA A100
 40 Gb/s Access Switches
 3 FAT Nodes  19 Edge Clusters
 39 NAS Nodes  VPN Remote Access

 
Scientific Cloud
The BDSC has, further, deployed a HPC scientific cloud, equipped with virtualization environments, making, locally and remotely, available all the scientific software used by the SSRF users:
Featured Software
 GISAS-2  PITRE  Aquarium
 Nautilus  RAW  Axis 2000
 Demeter  MAUD  SOLIDWORKS
 Ansys  Avizo  CT Reconstruction Packages
 MATLAB  Igor Pro  Containers
 MongoDB  SciCat  Finback
 TensorFlow  PyTorch  cuDNN
 LAMMPS  CUDA  Java
 Python  C / C++  ......
 
Moreover, the BDSC has implemented and integrated a unified authentication system within its centralized AI-SSRF-SP, therefore the SSRF users can now log in to the BDSC AI-SSRF-SP using their Large Research Infrastructures User Service Platform (CAS LSSF), SSRF or Shanghai Advanced Research Institute (SARI) credentials, and augment their Science with the next-generation Scientific Computing. 
 
Contact

附件: