The Department of Energy, National Science Foundation, and the National Aeronautical Space Agency support an ecosystem of “leadership class” computing facilities housing some of the world’s most advanced supercomputers and high-end visualization and data analysis resources. These facilities provide “free” computing cycles and storage to researchers from academia. Access to these resources is obtained through an application process that is based on the merit of the research objectives, and demonstration of the efficacy and parallel scalability of the software to utilize such computing resources.
The aim of the IDRE Pipeline Program is to inform UCLA researchers about the capabilities of these free national leadership class facilities, to facilitate UCLA researchers in obtaining allocations, and if needed, to scale their software to run on many more compute cores to advance their research objectives. The intentions are also to develop a community of researchers who want to Pipeline their software from running on Hoffman2 to leadership class facilities as well as Pipeline their software from running on their desktop to Hoffman2. The Pipeline Program will facilitate faculty working together to solve software bottlenecks and for faculty to work with IDRE-RTG staff.
This program includes regular seminars, support for preparing competitive proposals to the National Leadership Class Computers, as well as a quarterly call for proposals for IDRE-RTG staff to work closely with a research group to Pipeline their software to run an order of magnitude more effectively. For example, Hoffman2 currently has 10,000 compute cores while the computers at the leadership class facilities have nearly 1,000,000 cores. Therefore these facilities can be used to tackle much bigger or much more challenging problems than Hoffman2. The Pipeline Program also includes making or providing tools for groups to profile their codes on Hoffman2 in order to identify bottlenecks in performance.
A list of the current facilities along with respective deadlines for submitting allocation requests can be found at additional computing resources.