TY - JOUR
T1 - Translational research in the MPICH project
AU - Gropp, William
AU - Thakur, Rajeev
AU - Balaji, Pavan
N1 - Funding Information:
Some of the roadblocks for translational research have not always been present, for example, when MPICH (and the MPI standardization effort) started. Perhaps the biggest challenge has been a change in the timeline for funding, which is now mostly built around short-term (3–5 year) grants and projects. This approach may work if a team has the ideas ready and only needs to implement them, but if the team needs a few years to develop the ideas and iterate with the user communities, this time is not enough. It is even more difficult if the potential is unclear, as was the case for the MPI standard process, which was not the first attempt to create a message passing standard. Another challenge is the artificial separation between basic research and translational research. We believe that the best translational software involves both basic and applied researchers from the beginning through implementation, deployment, and future developments. Finally, no practical mechanism exists to sustain open source software, or, in other words, there is no general business model for supporting the software. Software needs to be maintained and often reimplemented as both the hardware and the system software changes. Someone must bear those costs, but most users simply see open-source software as “free” and do not recognize the costs of providing, maintaining, updating, and optimizing the software.
Publisher Copyright:
© 2020 Elsevier B.V.
PY - 2021/5
Y1 - 2021/5
N2 - The MPICH project is an example of translational research in computer science before that term was well known or even coined. The project began in 1992 as an effort to develop a portable, high-performance implementation of the emerging Message-Passing Interface (MPI) Standard. It has enabled the widespread adoption of MPI as a way to write scalable parallel applications on systems of all sizes including upcoming exascale supercomputers. In this paper, we describe how the translational research process was used in MPICH, how that led to its success, the challenges encountered and lessons learned, and how the process could be applied to other similar projects.
AB - The MPICH project is an example of translational research in computer science before that term was well known or even coined. The project began in 1992 as an effort to develop a portable, high-performance implementation of the emerging Message-Passing Interface (MPI) Standard. It has enabled the widespread adoption of MPI as a way to write scalable parallel applications on systems of all sizes including upcoming exascale supercomputers. In this paper, we describe how the translational research process was used in MPICH, how that led to its success, the challenges encountered and lessons learned, and how the process could be applied to other similar projects.
KW - MPI
KW - MPICH
KW - Message passing libraries
KW - Translational computer science
UR - http://www.scopus.com/inward/record.url?scp=85090062140&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85090062140&partnerID=8YFLogxK
U2 - 10.1016/j.jocs.2020.101203
DO - 10.1016/j.jocs.2020.101203
M3 - Article
AN - SCOPUS:85090062140
VL - 52
JO - Journal of Computational Science
JF - Journal of Computational Science
SN - 1877-7503
M1 - 101203
ER -