Meet NWChemEx’s Stakeholders

NWChemEx is being developed for an audience. Who is that audience? What do they want? How do we give it to them? That’s the point of this page.

Executive Summary

The key design points from this analysis are:

  • Performance (needed to make running on supercomputers worthwhile)
  • Ease of use (too high of a barrier will inhibit use cases where there are other choices)
  • Library instead of program mentality (we shouldn’t assume we are driving the workflow)
  • Extensibility (adding new theories should be easy)
  • Varying levels of programming expertise (research is done by undergraduates, graduates, post docs, scientists/professors from a multitude of backgrounds)

Stakeholders

Note: The stakeholder categories are NOT mutually exclusive. For example, most authors will also be end users. The point of the categories is not to group actual people, but rather to deduce design decisions based on the needs of each category.

Stakeholders are people that will interact with NWChemEx either directly or indirectly. The key stakeholders we have considered are:

  • End Users: People who use NWChemEx to perform electronic structure calculations.

    • E.g., scientists, researchers, students, etc. trying to understand the mechanism of a chemical reaction by running a series of calculations.
    • May or may not be programmers.

  • External Developers: People who develop software, but do not target NWChemEx during that development.

    • E.g., developers of math libraries, communication libraries.
    • Also includes developers of other electronic structure packages.

  • Contributors: People who develop software that targets NWChemEx, but is not housed by the NWChemEx GitHub organization.

    • E.g., Plugins that live outside the NWChemEx organization, tools designed to work with multiple electronic structure packages.
    • Upstream contributors are developers want NWChemEx to call their software.
    • Downstream developers want to call NWChemEx in their software.
    • We have no formal definition of “targeting”. If a developer wants to be included in the contributor category, they are.

  • Authors: People who contribute directly to repositories housed by the NWChemEx organization.

    • E.g., developers of SimDE or SCF.
    • Any contribution counts, from documentation to code.
    • Again up to the developer if they want to be included in this category.

  • Beneficiary institutions: Institutions that benefit from NWChemEx. These institutions may or may not have people that directly interact with NWChemEx. Examples include universities, national labs, research institutes, etc.

    • E.g., U.S. Department of Energy, Ames National Laboratory, Iowa State University.

Use Cases by Stakeholder

End Users

End users want to run calculations and generate data. They are unlikely to be concerned with how NWChemEx is implemented, rather they care about what it can do. Uses:

  • Computing properties of a molecular system.

    • E.g., energy, dipole moment, vibrational frequencies, etc.
    • These calculations are done in small batches.
    • For very small batches users may prepare inputs by hand.
    • For larger batches users may use workflow software to prepare inputs.

  • High-throughput calculations.

    • E.g., screening a large number of molecules for a desired property.
    • These calculations are done in very large batches, to the point that preparation of traditional input files would be expensive.

  • Routine and heroic calculations.

    • Not all calculations will be on the largest supercomputers, but some will.

External Developers

External developers are not too concerned with NWChemEx itself, rather the use cases here center around how contributors and authors can leverage other community software to make NWChemEx better. Use cases:

  • Ability to leverage existing high-performance libraries.
  • Contribute back to upstream libraries to improve performance and/or capabilities.

Contributors

Contributors are concerned with leveraging NWChemEx. Use cases are:

  • Integrate their software seamlessly into NWChemEx.
  • Integrate NWChemEx seamlessly into another software package.
  • Reentrant calls to NWChemEx.
  • To offload high-performance algorithms to NWChemEx. This facilitates adopting their software to funding proposals looking for high-performance applications.

Authors

Authors primarily want to develop new theories, algorithms, and methods. Use cases are:

  • Developing new electronic structure methods.
  • Developing new properties.
  • Developing new analysis techniques.
  • Adapting algorithms to new hardware.

It is important to remember that computer scientists and applied mathematicians need software frameworks for their research too, i.e., authors are not just chemists and physicists.

Beneficiary Institutions

Beneficiary institutions want to see NWChemEx benefit their missions.

  • A tool for funding. When funding calls are amenable to computational chemistry, researchers at the institution can leverage NWChemEx to write a convincing proposal.
  • Exemplar of a software portfolio, e.g., be able to say things like NWChemEx is used for X% of all quantum chemistry calculations on DOE supercomputers.
  • As recruitment aid, e.g., if you want to do computational chemistry at institution Y, you can be a part of the NWChemEx community.

Expectations by Stakeholder

End Users

End user expectations are focused on the experience of using NWChemEx. Key expectations below should be EASILY accomplished:

  • Able to run anywhere from one to a “very large number” of calculations.
  • Can obtain good performance on all platforms.
  • Accurate results.
  • Good documentation and examples.
  • Active support community.

The emphasis on “easily” is because being able to do the above is not enough; users don’t want to hunt for answers or spend hours tuning the software. They want an experience that “just works.”

Developers

TODO

Authors

  • Rapid prototyping. New development should be as quick as possible and ideally have some semblance of performance.

Beneficiary Institutions

TODO

Design Implications

With the above use cases and expectations in mind, we can now map those back to design implications for NWChemEx.

Performance Focused

NWChem fulfilled a niche in the community by being able to run on the largest supercomputers. With NWChem being deprecated, NWChemEx has the opportunity to pick up the mantle. The broader community needs software targeted at large runs. By focusing on performance we address the following use cases/expectations:

  • End users wanting to run heroic calculations.
  • Easily adopt community libraries that provide high-performance capabilities without loosing the benefits of those libraries.
  • Contributors wanting to their software to benefit from NWChemEx’s performance.
  • Ability for authors to quickly prototype performant algorithms on new hardware.
  • “Bragging rights” for beneficiary institutions.