Freva is dead, long live Freva! New features of a software framework for the Earth System community

Freva (the Free Evaluation System Framework [1; 2]) is a platform developed by the earth science community for the earth science community. Designed to work over HPC environments, it efficiently handles the data search and analysis of large projects, institutes or universities. Written on python, the framework has undergone a major update of the core. Freva offers:

• A centralized access. Freva comes in three different flavours with similar functionalities: a command line interface, a web user interface, and a python module that allows the usage of Freva in python environments, like jupyter notebooks.
• A standardized data search. Freva allows for a quick and intuitive search of several datasets stored centrally. The datasets are internally indexed in a SOLR server with an implemented metadata system that satisfies the international standards provided by the Earth System Grid Federation.
• Flexible analysis. Freva provides a common interface for user defined data analysis tools to plug them in to the system irrespective of the used language. Each plugin can be encapsulated in a personalized conda environment, facilitating the reproducibility and portability to any other Freva instance. These plugins are able to search from and integrate own results back to the database, enabling an ecosystem of different tools. This environment fosters the interchange of results and ideas between researchers, and the collaboration between users and plugin developers alike.
• Transparent and reproducible results. The analysis history and parameter configuration (including tool and system Git versioning) of every plugin run is stored in a MariaDB database. Any analysis configuration and result can be consulted and shared among the scientists, offering traceability in line with FAIR data principles, and optimizing the usage of computational and storage resources.

 

Freva has also experienced an upgrade on the sysadmin side:

• Painless deployment via Ansible, with a highly customizable configuration of the services via Docker.
• Secure system configuration via Vault integration.
• Straightforward migration from old Freva database servers or between Freva instances.
• Improvements in the dataset incorporation.
• Automatic backup of database and SOLR services.

[1] https://www.freva.dkrz.de/
[2] https://github.com/FREVA-CLINT/freva-deployment