Cloud-native geo: Practical lessons

Lukas Merz shares practical lessons from real-world projects on the #cloudnativegeo paradigm: what it can do, what not, and when PostGIS (or similar technology) is simply enough. In a second article, he walks through a concrete implementation using InSAR data in a web platform.
Author

Ralph Straumann

Published

April 27, 2026

Cloud-native1 geo(spatial) is a hot topic.2 There is certainly also some hype and inflated hopes around it. Recently, Lukas Merz3 has published two, in my opinion very noteworthy, articles grounding the #cloudnativegeo paradigm with practical lessons learned from real-world projects:

Cloud-native geospatial: What it can do, what not – and when you really need it: This article briefly explains the cloud-native and the cloud-native geo paradigms, then briefly covers important cloud-native geo formats, and lists (high-level) success factors and pitfalls.

Excerpt:

Our recommendation: Don’t ask, “How do we make our data cloud-native?” but rather, “What problem are we trying to solve? And is a cloud-native format the best solution?” Sometimes the answer is “yes.” Sometimes it’s “PostGIS is enough.” Both are good answers, provided they’re based on a sound assessment.

Cloud-native geospatial in practice: InSAR data on the GIN natural hazards platform: This article covers how cloud-native geo paradigm was used in a real-world project, the GIN natural hazards platform, to make InSAR data available and usable within a web application. It covers the technical implementation, problems the team encountered, and how it overcame them.

Excerpt:

Integrating the InSAR data into GIN was an instructive project. The combination of COG, PMTiles and Parquet/DuckDB worked well. It worked best where the conditions were right: large, static datasets, simple access patterns, and an update pattern that fits perfectly with file-based formats. At the same time, the project demonstrated that ‘cloud-native’ is not a one-size-fits-all solution. Each of these formats comes with its own optimisation requirements, and smart data preparation remains the key factor in performance.

Both articles are in German; in-browser translation works fine.

Footnotes

  1. I.e., taking full advantage of the cloud computing model; in the data world typically through enabling partial downloads of file contents through so-called HTTP range requests.↩︎

  2. For example, a quick search tells me it has been featured in this very blog in at least 9 articles.↩︎

  3. Transparency note: Until recently, Lukas was a colleague of mine.↩︎