Lessons learned and challenges in maintaining the ViaLactea knowledge base

The ViaLactea Knowledge Base (VLKB) was designed and initially developed within the EU FP7 VIALACTEA project that included a Work Package dedicated to create infrastructure and tools to perform research in Milky-Way astrophysics. The infrastructure’s goal was to set up an archive and services to enable that research. About 50 dataset collections ( 35k datasets of various sizes, in FITS format), ten catalogues of compact sources (from thousands to a few million rows), a catalogue of morphological complex sources (few thousand sources), and a few other catalogues and simulated datasets were included in the archive, worth about 1TB of data. On top of those data, and their metadata descriptions, a set of services was deployed: a search and access (cutout and merge) service for the datasets, a general Table Access Protocol (TAP) service for all metadata and catalogues and some other dedicated attempts for serving specific user requirements. All the interfaces were developed in combination with the dedicated client, the ViaLactea Visual Analytics (VLVA) but were designed keeping in mind the discovery and access scenario that is continuously developed in the Virtual Observatory (VO) ecosystem. Indeed, interoperability was brought inside the VLKB afterwards, slowly (depending on the limited resources available after the end of the Vialactea project), mostly when the VLKB resources kept being used in galactic astrophysics projects or as a comprehensive resource of data and services in technical demonstrator projects. Those projects provided the continuity in funding basic maintenance of the VLKB and some updates (even if occasionally rather than continuously). With the first release of the VLKB in 2016, the subsequent maintenance gap spanning from then until 2020, and the restart of development since then, the current adoption of standards in the VLKB includes: an ObsCore table to keep the metadata for the observational datasets’ catalogue, the TAP service to expose the general metadata content for all its data resources (catalogues, images, radial velocity cubes and morphological complex objects, . . . ), a custom implementation of the SODA (Server-side Operation for Data Access) standard set up to replace the dataset cutouts (with UWS - Universal Worker Service - used to manage asynchronous cutout and merge requests). Furthermore, authentication and authorization infrastructure (AAI) solutions using OAuth/OIDC have been tested on top of the cutout service, and a multi-cutout solution has been presented at VO level as a feedback for the SODA and DataLink evolution. Other features (management of complex morphology, of simulated data, and registration of the VLKB resources into the VO Registry) are still undergoing or missing. In particular, enabling more standard interface for the VLKB and making VLVA aware and able to consume them, will let both the client be more general and easier to maintain and the server resources be consumed by non-dedicated client applications. This contribution reports the challenges in maintaining and improving the VLKB, the actual status of the technologies and standards in use for its resources, and the present and future perspectives for the VLKB itself.