Director of Science
Jack Wells is the Director of Science for the Oak Ridge Leadership Computing Facility (OLCF), a Department of Energy (DOE), Office of Science national user facility, and the Titan and Summit supercomputers, located at Oak Ridge National Laboratory (ORNL). He is responsible for the scientific outcomes of the OLCF’s user programs. Jack is the newly elected vice chair of the American Physical Society Division of Computational Physics (DCOMP).
Jack previously led both ORNL’s Computational Materials Sciences group in the Computer Science and Mathematics Division and the Nanomaterials Theory Institute in the Center for Nanophase Materials Sciences. Prior to joining ORNL as a Wigner Fellow in 1997, Jack was a postdoctoral fellow within the Institute for Theoretical Atomic and Molecular Physics at the Harvard-Smithsonian Center for Astrophysics. Jack’s first experience at ORNL was in 1990 as a graduate student in full-time residence pursuing computational physics research and learning parallel computing from colleagues with whom he still works today at ORNL.
Jack has a B.S. in physics from Centre College and a Ph.D. in physics from Vanderbilt University. He has authored or co-authored over 100 scientific papers and edited 1 book, spanning nanoscience, materials science and engineering, nuclear and atomic physics computational science, applied mathematics, and text-based data analytics.
HPCwire: Hi Jack, congratulations on your selection as a 2019 HPCwire Person to Watch. Also congratulations on the impressive science already being tackled with Summit. Director of science at a leadership computing center sounds like the kind of title most HPCwire readers would love to have. What’s the actual job? What are your primary responsibilities and how have they changed (and are changing)?
Jack Wells: As the Director of Science for the Oak Ridge Leadership Computing Facility at Oak Ridge National Laboratory, I’m responsible for the scientific outcomes of our computing facility’s user programs. We provide leadership-class computing resources to researchers from government, academia, and industry who have many of the largest computing problems in science.
One can think of my job as an alliance manager creating synergy between our users and our leadership computing facility. I engage with our PIs, especially in our large programs (INCITE and ALCC), to ensure all aspects of our center are operated in a way that prioritizes their science outcomes. My primary responsibilities are in ensuring we are operating the facility in a way that prioritizes science; leading our team that manages our discretionary computing and data-storage resources and policies; engaging and requirements gathering for future procurements; and working with our PIs and science writers to develop materials that celebrate the accomplishments from our user programs.
My responsibilities haven’t really changed since I joined OLCF in 2011 as Director of Science, but some of the requirements and priorities have changed. Our users are demanding new capabilities to support research in data analytics and artificial intelligence. I work to help users to take advantage of new technologies to address their new needs, such as Summit’s tensor core GPU architecture. Additionally, we’ve had to change policies over time. One notable example concerns our policies supporting high throughput computing. Our leadership computing program has embraced policies supporting large-scale ensembles of calculations that cannot be effectively completed on smaller resources, rather than one huge simulation.
HPCwire: This question is a little in line with the last. Expectations for leadership class machines are understandably high; one has the sense they are not intended just to attack previously less tractable problems, but rather they are expected to add foundational insight to science domains (including computational science). What’s the mechanism and criteria for picking projects and judging success? How do you assess projects and how often are course adjustments needed where you say, we probably should have taken this particular project?
The mechanisms and criteria for project selection in our user programs are based on scientific and technical merit as determined by peer review. Researchers in any scientific discipline can apply for time on OLCF resources through several allocation programs. The INCITE program—our flagship program— promotes transformational advances in science and engineering for computationally and/or data intensive, large- scale research projects through large allocations of computer time and data resources. Awards are made based on the quality and impact of the research and the suitability of the proposed simulations for the requested resources. INCITE employs a two-phase review process. Proposals are peer-reviewed by a domain-specific panel of national academy-class experts for scientific and technical merit and potential impact. The computational readiness of the proposal is also assessed to determine how effectively the project will use the requested system.
HPCwire: Generally speaking, what trends and/or technologies in high-performance computing do you see as particularly relevant for the next five years? Also, what’s your take on near-term prospects for quantum computing and neuromorphic technologies?
Of course, I’m extremely excited about the science results that will emerge from Summit’s users. Data analytics and machine learning integrated in creative fashion with modeling and simulation is growing within our user programs. It is clear that hybrid, accelerated computing technologies are a large part of this story, and I’m excited to see how far these new approaches can progress on Summit. A growing feature is the use of mixed precision floating-point operations implemented in new hardware to achieve new heights in performance.
Also very important is the development of performance-portable programming tool chain for scientific computing on accelerated node architectures. This development includes, the evolution of language extensions and frameworks; maturation of accelerated compiler directives such as OpenACC and OpenMP; and the emergence and adoption of library abstractions, such as Kokkos and Raja.
The Exascale Computing Project is under full steam now, and it will be exciting to see the hardened, production-scale software technologies and applications that will emerge on the time scale of our exascale machine, Frontier. ECP just completed a very successful Annual Meeting in Houston in January, and the gathering was very impressive. This is the most significant investment in applications and software technologies in my career, and it is truly a national-scale enterprise.
HPCwire: Outside of the professional sphere, what can you tell us about yourself – personal life, family, background, hobbies, etc.? Is there anything about you your colleagues might be surprised to learn?
My wife, a physician working for the US Veterans Administration, and I have three kids in high school and a shelter dog. So, (big surprise!) life is very busy-fun. We enjoy cooking, travel, and hiking. I grew up in Prestonsburg KY, and I have the experience of working in coal mines during summer breaks during my college years.