Conclusions
The LROSE Science Gateway was developed to improve accessibility to weather radar and lidar analysis tools. By deploying JupyterHub servers on NSF Jetstream2 cloud resources, we have simplified the procedure for hosting educational opportunities for students and community members. The LROSE Science Gateway provides a variety of tutorials in a format widely-used by the community and works to balance ease of use with guided instruction.
Through use of containerization and Kubernetes we’ve installed JupyterHub and provided users of varying needs and experience levels with pre-configured environments in which to use LROSE tools, including rapid visualization tools which were not traditionally used from a browser-based interface. We’ve made use of multiple workflows to ensure efficient use of our allocation, and the team is currently exploring the use of one of Jetstream2’s newest developments, Openstack Magnum, to deploy true auto-scaling clusters.
The LROSE and NSF Unidata teams continue to explore how to make resources and tutorials most effective. We plan on continuing to evaluate the use of Openstack Magnum to provide users with a streamlined experience while accessing gateway resources. One potentially exciting use case is making use of Jetstream2’s A100 GPUs, which have historically been difficult to use due to their high SU consumption and the “idling resources” problem. Other areas of interest are incorporating CI/CD techniques to ensure gateway visitors have access to a bleeding edge “nightly” build of LROSE as well as a stable release. In terms of the tutorial content, our team notes areas of confusion during workshops, post-workshop surveys request feedback on the notebooks, and users can provide suggestions and critiques through GitHub issues or email. Ongoing efforts to organize more workshops, including more in-depth sessions targeting intermediate LROSE users, and the development of new tutorial notebooks requested by the community will continue.
We want to emphasize the importance of our cross-disciplinary collaboration. The partnership between CSU, NSF NCAR EOL, and NSF Unidata allows for efficient use of team members’ expertise. While the scientists in our team focused on evaluating software for correctness and developing new workflows to share with the community, the scientific software engineers worked on implementing new software features critical to advancing the state of the discipline, and the cloud software engineers aimed their efforts towards bringing LROSE to the gateway in a user-friendly way. Multiple advancements have been made or expedited as a result of this collaboration, including bug fixes, code performance enhancements, and feature developments such as virtual desktop technology.