...
| Idea Short Title | Submitter | Team Members | Idea Pitch/ Description/ Plan | Proposed Deliverable | Special Needs? | Notes |
|---|---|---|---|---|---|---|
| Spark-on-rhea | John Harney | John Harney, Ben Mayer, Valentine Anantharaj, Marcia Branstetter, Dan Ricciuto | OLCF is unveiling "spark-as-a-service" on Rhea. It is an elastic, on-demand type application that submits a batch script to Rhea that creates a spark cluster on the fly, reads data from lustre, performs calculations, and writes the result back to disk. This can be easily embedded into the larger ACME workflow. Not sure if there will be a need for it, but is there an analytics app that can be deployed within this framework? | A small-scale, working map-reduce solution/ application that can be embedded into the workflow in some way. | OLCF accounts? | Note: I'm not sure if this will be useful, but it may be interesting to explore this space within the context of the workflow. |
| Deep learning as-a-service | John Harney | John Harney, Ben Mayer, Valentine Anantharaj, Marcia Branstetter, Dan Ricciuto | Similar to the previous entry, OLCF is also trying to unveil "deep learning as a service" on OLCF (specifically Rhea). Deep Learning is a branch of machine learning that utilizes multi-layered neural networks to provide classifications of data products. The most common application in industry is image classification and speech recognition. Perhaps we can find a use case for this as well? | A small-scale working deep learning solution/application that can be embedded into the workflow in some way. | OLCF accounts? | Note: I'm not sure if this will be useful, but it may be interesting to explore this space within the context of the workflow. Also, deep learning is increasingly becoming a very popular ML technique. It may be crucial to get in on the ground floor. |
| Data management for ACME | Lukasz Lacinski | Lukasz Lacinski, Rachana Anathakrishnan | Transfer, sharing and publication of data uses various capabilities in ESGF built on Globus service. During this time, we'll provide hand-on help and training for ACME scientists and other workflow team members (e.g. process flow) on using such capability and integrating it with application as needed. | Documentation on using these capabilities | Depending on interest this could be purely end user focused for say transferring results from ALCF/NERSC (HPSS) to ORNL and sharing it, or API discussion for other developers. | |
| Land model point mode/UQ hands on | Dan Ricciuto | Dan Ricciuto, Khachik Sargsyan, Peter Thornton | Demonstrate python-based workflow for execution of point-mode simulations for ALM, which can be useful for diagnosis and evaluation of new model developments. We will also walk through the steps required to run a model ensemble and sensitivity analysis. We will take suggestions from users about possible new options, and implement these as time allows. | Improved workflow scripts and documentation for ALM point model and UQ | OLCF/NERSC accounts | Interested land modelers are welcome to suggest specific questions that we can target in this breakout. |
| CIME integration | Andreas Wilke | Robert Jacob, James Foucar, Andreas Wilke, Jason Sarich, Anshu Dubey | CIME4 is undergoing some changes, e.g. new configuration and scripts, which are improving usability and performance. Those changes have to be integrated and tested within ACME. | Creating a beta version of ACME and CIME4 | ||
| Jenkins Best Practices | Jayesh Krishna, Jim Foucar |