Gordon: A Data-Intensive Computing Research Experiences for Undergraduates (REU at the San Diego Supercomputer Center, University of California, San Diego
Application Deadline – May 18, 2012
Project Background: Application Performance Analysis and HPC System Deployment
This research project involves the identification of computationally intensive applications and their evaluation on the new Gordon HPC system. Benchmarking plays a crucial role in the Gordon project since it enables us to confirm that the processors, memory subsystems, networks, and file systems are all working together as expected. More importantly, the benchmarking of real user applications will help us to determine the classes of problems that can make best use of this unique architecture.
Work done to date indicates that Gordon should be particularly well suited for applications that have some combination of large memory or I/O requirements, limited concurrency, and irregular patterns of data access. What makes these results especially exciting is that applications with these characteristics generally do not run well on traditional distributed-memory machines.
The student will analyze and profile a variety of both standard benchmarks and application codes from domains including structural dynamics, bioinformatics, chemistry, graph theory, and astronomy. The results will be used to help new users determine if their research is a potentially good fit with Gordon, and they therefore likely candidates for allocations. The student projects will assist the national allocations committee in deciding which projects should be awarded computer time on Gordon. The student will have the opportunity to work closely with SDSC staff with a range of technical backgrounds (computer science, physics, chemistry, and engineering) and will be expected to actively participate in team meetings.
Required Student Coursework
SDSC invites an REU student to work with Dr. Pietro Cicotti to explore further impacts of the data-intensive research era on metrics of computational leadership. The student will create a benchmark to measure DMC, start their own “Top500 DMC List,” and measure NSF and other agencies’ machines with the metric.
The ideal REU student for this position should have completed coursework in Electrical Engineering, Computer Science, or a related field at a college or university in the United States. Priority will be given to students who have some programming background in C/C++, FORTRAN, Perl, or Python and have experience working in a UNIX or Linux environment. Work will be done under the direction of Dr. Robert Sinkovits, Gordon Applications Lead.
Mentor Profiles:
Dr. Robert Sinkovits holds a PhD in physics and has been a computational scientist at SDSC for almost 15 years. His work focuses primarily on the development of parallel and highly optimized applications in a variety of domains including computational fluid dynamics, astrophysics, quantum electrodynamics, high-energy physics, structural biology, mass spectroscopy, and molecular dynamics. He has authored or co-authored approximately 40 peer reviewed journal publications plus a number of book chapters and conference proceedings. He is the primary author of the software packages auto3dem and IHRSR++, which are used to determine the structures of icosahedral viruses and helical particles, respectively, from electron microscopy data.
Dr. Pietro Cicotti holds a PhD in computer science from UCSD and is a computational scientist at SDSC. He is a member of both the Performance Modeling and Characterization Lab (Lab (http://www.sdsc.edu/pmac/) and a member of the Gordon Applications team. His primary research interests include aspects of the design of large-scale hardware and software systems for HPC and scientific computing. He is author of Tarragon, a library for data-flow execution on distributed memory machines, and he contributed to the SuperLU_DIST package for sparse linear systems and the BOINC platform. Dr. Cicotti works with domain experts on the development and optimization of applications in a wide variety of scientific domains.
Internship Timeline and Stipend
Successful student applications will receive a stipend of $7500 for a ten-week, full time summer position with start date to be arranged upon selection. Ideally, the student should begin within a few weeks after completion of spring finals to allow sufficient time for completion of the REU project. Selected students may also opt to begin their experience during the academic year. Such arrangements will be coordinated with the mentor upon selection.
The Application Process
Interested applicants are asked to send a letter of interest, a resume and recommendation letter (strongly encouraged) to:
Dr. Diane Baxter
Education Director
San Diego Supercomputer Center
University of California, San Diego
9500 Gilman Drive
La Jolla, CA 92093-0505
The application deadline is May 18, 2012.
Questions can be sent directly to Dr. Pietro Cicotti at pcicotti@sdsc.edu.
