A Hybrid, Hybrid Approach to Data Analysis Using Kubernetes and Cloud Technologies

The data in the world of space science has exploded. In the past, the data from space probes was smaller compared to the software, so the data would be brought to the software. This has changed with improved sensor precision and higher data capture rates. Rather than moving the large data volumes to the software, a new approach has emerged: shifting the software to the data.

Over the past year, we’ve dived into using Docker and Kubernetes to be able to handle heliophysics processing software. We’ve been able to build out modular webapps to visualize science and engineering data by using containers on different systems to showcase the potential of using cloud technologies for data storage and processing.

The talk will discuss the work we’ve done to deploy these web apps in a hybrid Kubernetes environment. This setup combines in-house servers, a high-performance computing cluster for heavy-duty tasks, and AWS. I'll discuss a spectrum of deployment strategies from 100% using Kubernetes and Longhorn/S3 buckets to a regular setup involving technologies like Apache, C++, python, and others on bare-metal. This flexibility ensures progressive enhancements as one adds in cloud tools, while preserving the option for a traditional system setup to cater to varying program needs and budgets, whether for a CubeSat project or a long-term, far-reaching mission.

Scalability, a crucial aspect of modern data analysis, will also be covered. I'll share methods for using queues to manage distributable tasks, such as with our GeoViz application. Additionally, I'll introduce useful container management tools for managing dependencies and on-premises repositories.

While fully relying on the cloud has benefits and drawbacks, the advantages of mixing on-site and cloud-based software shouldn't be ignored. Data volumes grow, software advances, and the cloud computing landscape keeps developing. The evolutions of each of these aspects offers many chances to design adaptable, scalable, and efficient data analysis solutions to fit the ever-changing demands of space science exploration.