I and my team strive to enable scalable HPC data analytics of scientific data. In other words, we study the use of high-performance supercomputers for the analysis and visualization of scientific data, in addition to supercomputers' traditional role for simulation and modeling. The data we analyze originates in scientific instruments and computer simulations, and the supercomputers and science facilities are some of the largest in the world.
We study the problem from three perspectives: scalable algorithms, software infrastructure, and application engagement.
Scalable algorithms: Scalable data analytics requires efficient and parallel algorithms to handle high-volume and high-velocity data streams. The size and scope of the science problems we tackle and the machines required are well beyond commercial scale, and require research in new parallel algorithms that can scale to hundreds of thousands of processors and tens of millions of threads.
Software infrastructure: Libraries and other middleware for the rapid development of scalable analysis algorithms, and for their in situ coupling to simulations and experiments, is another area of active research. Programming models and runtimes for rapid development of individual parallel algorithms, parallel dataflow coupling, and workflow composition of multiple analyses with data sources are examples of software infrastructure.
Application engagement: We cannot solve the above problems without deep application engagement. Domains such as high-energy physics, materials science, and environmental science not only provide test cases to validate and test solutions, but drive our research in new directions. Much of our research has a strong interdisciplinary nature for this reason.
The team, which we have affectionately dubbed PEDAL (Parallel Extreme-Scale Data Analytics) consists of the following members:
Tom Peterka is a computer scientist at Argonne National Laboratory, scientist at the University of Chicago Consortium for Advanced Science and Engineering (CASE), adjunct assistant professor at the University of Illinois at Chicago, and fellow of the Northwestern Argonne Institute for Science and Engineering (NAISE). His research interests are in large-scale parallel in situ analysis of scientific data. Recipient of the 2017 DOE early career award and three best paper awards, Peterka has published in ACM SIGGRAPH, IEEE VR, IEEE TVCG, and ACM/IEEE SC, among other top conferences and journals. Peterka received his Ph.D. in computer science from the University of Illinois at Chicago in 2007, and he currently works actively in several DOE- and NSF-funded projects.