Invited talk (I10.1) André Offringa (ASTRON Netherlands Institute for astronomy)Theme: Delivering accessible and science-ready radio data
Designing radio-astronomical software for delivering science-ready productsModern radio observatories rely heavily on advanced software pipelines to produce scientific data. Generic radio observatories such as LOFAR and the future SKA, which (will) cater to a wide variety of science projects, need multiple pipelines that produce different product types. At the core of these pipelines are reusable "building blocks"; software components that perform operations such as calibration and imaging. While this division allows for reuse, compartmentalization and clear interfaces, even these isolated blocks by themselves can be extremely big and complex.
Both the pipelines and the software components are fundamentally difficult to develop, because they require a high level of domain knowledge from astronomers, deep technical knowledge about the instrument and an advanced level of software engineering to design large, complex software systems that process extremely large volumes of data. The higher the level of 'science-readyness' is desired, the more difficult this becomes. Moreover, it is often not possible to design the components and pipelines from the start, because they require pioneering development: a theoretical solution might not work under certain conditions, are be too slow in practice.
I will discuss these challenges, and present the lessons learned from designing several software packages and pipelines for generating advanced radio data products.