Facilities
Computing Facility
Our state-of-the-art computing facility provides researchers with a 64-node Apple G5 compute cluster and a 9TB BlueArc file server.
We have installed a load balancing system so that upon login users are routed to a node under low load. From there users can run interactive neuroimaging applications like AFNI, FSL, FreeSurfer, and general purpose packages like Matlab. Users can also submit scripts using these and other packages to be run on dedicated nodes reserved for batch jobs. Using a simple interface, batch job progress can be monitored and email received when the job completes. Large datasets can be processed either on single nodes running 64-bit versions of these popular packages, or by splitting the data into pieces (such as image slices) for parallel processing.
The closely coupled file server is visible to the scanner, mounted on all cluster nodes, and on desktops throughout the building as well via high speed ethernet links. This arrangement allows raw data to be sent from the scanner to the file server, processed on the cluster, and the results visualized on desktops, using a single, secure host for data management.
Benjamin Singer
email: bdsinger@princeton.edu