Sept. 9, 2020 — This summer four Cornell University students learned about the latest cloud computing technologies and made valuable contributions to the Aristotle Cloud Federation and the computational tools researchers use to make scientific breakthroughs. Their work and learning experiences were made possible by funding from the National Science Foundation Research Experiences for Undergraduates (REU) program.
REU student Matthew Farnese improved the pipeline used to identify Fast Radio Bursts (FRB) by testing the functionality of the Friends-of-Friends algorithm, extending the existing pipeline methods for use with multiple data formats, creating new plotting methods, and adding features to make the pipeline more user-friendly. He also made progress towards a second FRB detection with the Friends-of-Friends algorithm in the Breakthrough Listen dataset that scientists use to search for evidence of intelligent life beyond earth. Five Cornellians worked with him on the project: Astronomy Professor Jim Cordes, Shami Chatterjee, Akshay Suresh, Peter Vaillancourt, and former Aristotle REU student Plato Deliyannis.
“Now that I have an idea of what computational astronomy/physics is like, I am even more interested in it, and I am considering getting my PhD in computational physics,” says Farnese. “In 10 years, I hope to be working at a university, whether it be as a lecturer, post-doc, or professor.”
The initial goal of Cornell REU student Priyanka Dilip’s research was to provide a GPU computing environment that scientists could use to install and test their applications. Over the course of the summer, multiple research groups from within the Aristotle Cloud Federation, as well as Cornell researchers, reached out to her to request specific cloud computing images and Virtual Machine instances, thus her research question became “How can Machine Learning frameworks and/or computational fluid dynamics (CFD) solvers be provisioned reliably as a GPU-accelerated resource?”
In response, Dilip created Cornell Red Cloud images containing TensorFlow, MATLAB, PyTorch, and Jupyter which were developed in an Anaconda environment and Dockerized for Forest Large Eddy Simulations and cryo-electron microscopy. She also containerized an OpenFOAM CFD solution, developed Dockerfile and images for RapidCFD, created GitHub test applications for Tensorflow+Keras and PyTorch, and wrote documentation to help others create server images and use GPUs more effectively.
“I definitely accomplished more than I expected,” says Dilip, “and virtual conferencing with Aristotle cloud systems engineer Bennett Wineholt every other day helped me make more efficient progress and clear doubts rapidly.” Dilip’s documentation will be published on the Cornell Center for Advanced Computing Wiki for Aristotle and Red Cloud users and the research community. After her REU experience, Dilip was pleased to learn she was selected by NVIDIA’s Deep Learning Institute to attend their “Deep Learning and GPU Programming Workshop.”
Cornell undergraduate student Jeffrey Lantz wrote a guide on how to get started with Kubernetes, updated a Terraform-Ansible tool, and created a Terraform-Kubernetes tool. He then used High Performance LINPACK benchmarks to compare the cost and efficiency of the two tools. Lantz concluded that the Terraform-Kubernetes tool is faster to deploy and less costly to deploy as a cloud computing cluster. “An easy-to-use Terraform-Kubernetes tool that is MPI capable is something new,” he notes. “It can be used in nearly any field of study and gives scientists access to large compute resources at a relatively low cost.”
“I think having regular one-on-one meetings and group meetings, an interesting project, and a strategy of announcing what I planned on working on at the beginning of each day all helped my productivity,” Lantz says. He also suggests that the REU program could be improved by providing students early on with “research tips” such as “research rarely works on the first try” and “don’t be discouraged if you’re having trouble with something or if it takes longer than you expect.”
Sherri Tan worked primarily under Professor Sara C. Pryor, a Cornell professor in Earth and Atmospheric Sciences and Aristotle use case scientist. The goal of her project was to predict the occurrence and magnitude of wind gusts surrounding three major airports: Newark, Boston, and Chicago. She proceeded to download datasets of predictors such as upper air variables using Python scripts, and then aligned those with datasets for the observed occurrence and magnitude of wind gusts at the airports using MATLAB. Tan then used MATLAB’s functions and toolboxes for generalized linear regression, stepwise regression, and deep learning to build predictive models and calculate descriptive statistics. She observed that Aristotle cloud resources enabled the download and efficient processing of large amounts of data.
“The REU experience gave me the opportunity to learn how to write scientific reports in order to document work and make research reproducible,” Tans says. “It also helped me build on my MATLAB programming and data science skills.”
“I’m grateful that my summer REU experience has provided me with greater insight and clarity on which direction I would like to pursue in the future and want to thank my mentor Professor Pryor as well as the Center for Advanced Computing’s REU lead Adam Brazier,” she says.
About the Aristotle Cloud Federation
The Aristotle Cloud Federation project is supported by National Science Foundation grant number OAC-1541215 and the Cornell University Center for Advanced Computing (CAC) in partnership with the University at Buffalo Center for Computational Research (CCR) and the University of California, Santa Barbara Department of Computer Science. UCSB undergraduate Kerem Celik also participated in this summer’s REU developing a telemetry data visualizer for the Citrus Under Protective Screening project and the Edible Campus farm.
About the Research Experiences for Undergraduates (REU)
The Research Experiences for Undergraduates (REU) program supports active research participation by undergraduate students in any of the areas of research funded by the National Science Foundation.
Source: Cornell University Center for Advanced Computing
