Külli Ek
Geoinformatics specialist at CSC - IT center for science (Finland):
* Geoinformatics support for Finnish universities and research institutes to use CSC supercomputers and cloud services
* Develoment of Paituli spatial data download service, inc STAC.
Sessions
Geoinformatics applications often include analyzing high volumes of data which may require a lot of time and computing resources. Many of these applications can benefit from high performance computing resources (supercomputers) to speed up the computation, or even make them possible -more memory, storage space, available tools and a computing system suitable to handle big data. They also provide more processing units (CPU and GPU) than an average research computer, which are essential components for efficient computational analysis. Particularly Deep Learning applications benefit from the use of one or multiple GPUs.
One of these supercomputers is LUMI, provided by EuroHPC Joint Undertaking and 10 European consortium countries. LUMI is particularly well suited for large scale modeling and deep learning applications. LUMI supercomputer is available for free for European academic researchers and for companies and public organizations for open R&D purposes.
Compared to commercial computing options, where technical support is rather limited, CSC and LUMI partners offer case-by-case support for projects. Also a wide range of courses is provided to get familiar with supercomputers.
This presentation aims to introduce the audience to supercomputing for geoinformatics tasks as well as the benefits and challenges that a move to the supercomputer may introduce for researchers and companies. It will also highlight some of the recent use cases from geoinformatics.
In this presentation we discuss our experiences from setting up Paituli STAC, which contains open Finnish raster datasets. We did not do any software development, but decided to use GeoServer with STAC extension. Own code was only written for populating the PostGIS database with information about ~100 collections and ~250 000 items. Paituli STAC catalog is mainly targeted for data analysis use cases, but can be used also from web applications.
We have also prepared public example scripts for using Paituli STAC with Python and R. We will also show the results of some scaling tests of using data from STAC on a supercomputer with Dask and xarray.
More information: https://paituli.csc.fi/stac.html