Satellite-based Data Lakes: A technical Case Study based on existing Cloud Technologies

Paper number

IAC-24,B5,1,5,x87639

Author

Dr. Markus Sauer, Siemens AG, Germany

Coauthor

Dr. Alessio Netti, Siemens AG, Germany

Coauthor

Ms. Elizaveta Boriskova, Siemens AG, Germany

Coauthor

Dr. Florian Zeiger, Siemens AG, Germany

Year

2024

Abstract

The deployment of satellite mega-constellations providing broadband Internet services has become more affordable than ever, thanks to the development of novel technologies such as Inter Satellite Links (ISL) and lightweight Low Earth Orbit (LEO) payloads. Coupled with the adoption of inexpensive but still powerful off-the-shelf hardware, this has renewed interest towards the use of satellite networks for computational tasks, similarly to traditional Edge or Cloud environments - this paradigm is commonly known as Satellite Computing. On-board satellite computation could be leveraged in a wide range of applications (e.g., earth observation, environmental monitoring or emergency response, and many more industrial applications), enabling them to operate holistically on the Earth and to serve requests anywhere and anytime. 
However, while satellite hardware is approaching the required readiness level, there is little insight on the suitability of common Cloud components, such as orchestration and data lake frameworks, for applications of this kind.
With this work we tackle this area of research, proposing a case study about the behavior of a distributed database deployed on a highly-mobile LEO satellite constellation. In particular, we leverage on Apache Cassandra, a popular database implementation used in many large-scale deployments, and focused on Iridium NEXT as our reference satellite topology. To do so, we conduct a series of experiments using an in-house testbed for a satellite communication and computation, which allows us to simulate distinct scenarios over the chosen satellite network and deploy any distributed application. This way, we can observe the interaction between ground terminals (i.e., client applications) and satellites (i.e., Cassandra nodes) under a wide range of database schema, scaling and pressure configurations, and characterize the resulting performance patterns, resulting mostly from satellite mobility, as well as Cassandra’s architecture as a distributed database. Our results are promising, showing that an implementation such as Cassandra can be made suitable for large-scale satellite deployments given a few key elements, including efficient schema design, data locality and a satellite-aware gossip protocol.

Abstract document

IAC-24,B5,1,5,x87639.brief.pdf

Manuscript document

IAC-24,B5,1,5,x87639.pdf (🔒 authorized access only).

To get the manuscript, please contact IAF Secretariat.