2018 IEEE International Conference on Electro/Information Technology (EIT)
Numerical reproducibility has received increased emphasis in the scientific community. One reason that makes scientific research difficult to repeat is that different computing platforms calculate mathematical operations differently. Software containers have been shown to improve reproducibility in some instances and provide a convenient way to deploy applications in a variety of computing environments. However, there are software patterns or idioms that produce inconsistent results because mathematical operations are performed in different orders in different environments resulting in reproducibility errors. The performance of software in containers and the performance of software that improves numeric reproducibility may be of concern for some scientists. An existing algorithm for reproducible sum reduction was implemented, the runtime performance of this implementation was found to be between 0.3x and 0.5x the speed of the non-reproducible sum reduction. Finally, to evaluate the impact of using a container on performance, the runtime performance of the WRF (Weather Research Forecasting) package was tested and found to be 0.98x of the performance in a native Linux environment.
© © 20xx IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.
Sara Faraji Jalal Apostal, David Apostal, and Ronald Marsh. "Containers and Reproducibility in Scientific Research" (2018). Computer Science Faculty Publications. 14.