Computing

PNI is committed to maintaining leading-edge data analysis and management systems which help support the creation of novel analytical and data collection methods. PNI’s information technology infrastructure has been designed to support research methods requiring extremely high-performance processing coupled with low-latency, high-speed connectivity, such as the integration of fMRI scanners with HPC clusters, enabling real-time analysis of fMRI data with complex algorithms including Full Correlation Matrix Analysis.

Central File Server

Collection and analysis of data is central to Institute research, and the need for a centralized, reliable, shared storage pool is critical. The central file server provides 400TB of usable storage for research data, analyses, and administrative data.

Data from the file server is accessible on the PNI compute cluster and from user workstations both on campus and off. This greatly facilitates collaboration and the sharing of data, and eliminates the need for multiple copies of datasets, saving time and disk space. A 10Gb private network connects the file server to the compute cluster, while 1Gb connectivity is provided to user workstations for desktop analyses and data visualization.

PNI file sever.PNI file sever.

General Purpose Compute Cluster

Directly connected to the file server is a compute cluster has a total of 52 nodes. Each node contains two quad-core Xeon X5570 processors (Intel), operating at 2.7GHz, for a total of 416 processors, and 24GB of RAM for a total of 1.25TB RAM. The nodes communicate with each other via DDR (20Gbps) Infiniband (Mellanox) in a flat, nonblocking network.

NSF-Intel High Performance Computing Cluster

The Princeton Full Correlation Matrix Analysis (FCMA) Toolbox is a suite of tools for correlating, analyzing, and classifying patterns of correlation in fMRI time-series data. Due to the enormous computational demands of the FCMA project, a high-performance compute cluster has been built to carry out the work. The cluster has been built with funds from the National Science Foundation (Major Research Instrumentation award 1229597). The cluster contains 6,000 coprocessor cores, from a combination of Xeon Phi boards and Sandy Bridge CPUs donated by Intel.