NERSCPowering Scientific Discovery Since 1974

Early Career Achievement Award Seminar Series


NERSC is hosting an online seminar series featuring talks from, and discussions with, the recipients of the NERSC Achievement Awards for early career scientists. The speakers will give a description of their research and significant results, describe their computational methods and/or strategies, and relate notable HPC challenges or successes at NERSC. They will also share their thoughts on what it’s like to be an early career computational scientist in today's environment.

The talks are open to anyone, see "Connection Information" below.


DatePresenterTitleTime (Pacific)
September 22  Antonio Villarreal, Argonne National Laboratory LSST DESC Second Data Challenge (DC2) Image Simulation Campaign with Parallel Python Workflows  12:00
September 29  David Vartanyan, University of California, Berkeley  TBA  12:00
October 6 Miha Muskinja, Lawrence Berkeley National Laboratory Raythena: A Massively Parallel Data Processing Framework for the ATLAS Geant4 Simulation  12:00
October 13  Grant Johnson, Princeton University & Lawrence Livermore National Laboratory  TBA  12:00
October 20 Quentin Riffard, Lawrence Berkeley National Laboratory TBA  
October 27 Hsin-Yu Ko, Cornell University Towards an Accurate and Efficient Order-N HPC Framework for Large-Scale Condensed-Phase Hybrid Density Functional Theory  12:00
November 3 Abigail Polin, Caltech & Carnegie Observatories  TBA  12:00
November 10 Samuel Kachuck, University of Michigan TBA  12:00

Connection Information

  • Berkeley Lab employees and affiliates: the ZOOM info is on the "NERSC Public Events" calendar. 
  • NERSC Users: See your NERSC weekly email or  this page.
  • General public: Please register.


LSST DESC Second Data Challenge (DC2) Image Simulation Campaign with Parallel Python Workflows

 Antonio Villarreal, Argonne National Laboratory

September 22, 2021
12:00-1:00 Pacific Time

The Vera Rubin Observatory LSST is going to provide the astrophysics community with an unprecedented amount of survey data with which to contain the evolution of the universe through time. In order to leverage this dataset, we will ultimately require extensive simulations in order to validate scientific pipelines ahead of the survey ever seeing light. The LSST Dark Energy Science Collaboration (DESC) Second Data Challenge (DC2) represents the largest simulated sky survey of its complexity. Generating such a simulation required managing a complicated and rapidly changing workflow across multiple compute resources. We demonstrate how we utilize containerization and the Parsl parallel scripting library in order to create a portable and scalable workflow to meet the challenges of this computational task. With this workflow we were able to generate a simulated survey volume covering 300 square degrees and five years of image depth, utilizing 100M hours of compute and up to 2000 Cori KNL nodes at a time. We discuss possible improvements that could be made to the workflow for future survey simulation, both from the standpoint of utilizing the increasingly common workflow nodes at high performance computing (HPC) centers and that of how the underlying image simulation code may be altered to benefit more from computing at these scales.

Raythena: A Massively Parallel Data Processing Framework for the ATLAS Geant4 Simulation

Miha Muskinja, Lawrence Berkeley National Laboratory

October 6, 2021
12:00-1:00 Pacific Time

Raythena utilizes the Ray software (a high-performance distributed execution framework) to distribute the highly intensive ATLAS Geant4 simulation workflow across a few hundred HPC nodes. Geant4 simulation is the most computationally expensive step of the ATLAS Monte Carlo simulation chain and represents about 50% of the ATLAS computing budget. Conventionally, it is run on ‘grid’ sites and each simulation campaign takes a few months to simulate the desired quantity of proton-proton collision events. Raythena is a solution for running ATLAS Geant4 simulation efficiencly on HPCs and it could significantly reduce the duration of simulation campaigns in the future. The goal of Raythena is to process as many events as possible with a given CPU-hour allocation on an HPC as fast as possible. An effective mode of operation at NERSC’s Cori was found to be running 100-200 Cori KNL node jobs in the flex queue. Raythena is a central application that orchestrates the workload management across all nodes using the Ray API. On Cori KNL nodes, 132 Geant4 processes were spawned on each compute node, amounting to more than 25,000 Geant4 processes running in parallel. Raythena handles communication with the ATLAS central PanDA database where it retrieves the input events and feeds them to the Geant4 processes. The Raythena framework was found to scale very well up to 100 to 200 nodes on Cori KNL with virtually no delay between the consecutive processed events.

Towards an Accurate and Efficient Order-N HPC Framework for Large-Scale Condensed-Phase Hybrid Density Functional Theory

Hsin-Yu Ko, Cornell University

October 27, 2021
12:00-1:00 Pacific Time

By including a fraction of exact exchange (EXX), hybrid functionals reduce the self-interaction error in semi-local density functional theory (DFT), thereby providing a more accurate and reliable description of the electronic structure in systems throughout chemistry, physics, and materials science. However, the high computational cost associated with hybrid DFT limits its applicability when treating large-scale and complex condensed-phase systems in many practical applications (e.g., design of fuel cells). To overcome this limitation, we have devised a highly accurate and linear-scaling (order-N) approach based on a local (e.g., MLWF) representation of the occupied space that exploits sparsity when evaluating the EXX interaction in real space. Powered by NERSC resources over the past several years, our development has evolved into a general-purpose algorithmic framework (exx) capable of efficiently using several modern HPC architectures [1]. The exx code already enabled several large-scale hybrid DFT-based applications in its pilot forms, e.g., unraveling the qualitative/substantial difference between the structural diffusions of H3O+ and OH- in aqueous solutions. Recently, we further extended exx to a black-box solver by eliminating its system-dependent parameters. With this new extension, exx brings us one step closer to the routine use of more reliable hybrid DFT for studying large-scale condensed-phase systems relevant to energy sciences and beyond.

[1] J Chem Theory Comput 16, 3757 (2020).