Blue Waters at NCSA
Blue Waters Uses Globus for File Sharing and Transfer
The Blue Waters team at the National Center for Supercomputing Applications utilizes Globus sharing so users can quickly, effectively, and securely share data sets with their research community or the broader public. There are two classes of sharing based on the needs of the partners and data: active data sharing for projects with current allocations on Blue Waters, and community sharing for data produced by prior projects. Users from each group can share data using Globus sharing capabilities via an easy to use web service interface.
In just one recent example, as part of this groundbreaking project pioneering the use of GPU-accelerated deep learning for rapid detection and characterization of gravitational waves, scientists ran numerical relativity simulations on Blue Waters that they then archived using Globus; they also used Globus to transfer the extracted simulations, which were obtained using POWER on Blue Waters, to the GPUs they used to train and test the neural nets. NVIDIA and Wolfram Research also collaborated on this project.
In another recent example, scientists at the Soltesz Lab at Stanford University used Globus in a project to better understand epilepsy, reporting that they use Globus quite a lot for transferring data between their lab computers and Blue Waters, as well as for moving data between Blue Waters and XSEDE supercomputers such as SDSC Comet, where they run some of their simulations. Scientists report that they have found the Globus data transfer capabilities to be indispensable in these projects.