Poster Abstract

P.5 Angus Comrie (Inter-University Institute for Data Intensive Astronomy)

Efficient data processing for large image cube visualization

The CARTA software package allows users to visualize large data sets stored on remote servers. Large radio-astronomy images in FITS, HDF5 and CASA formats can be visualized efficiently in a browser-based front-end, without needing to retrieve data multiple times when rendering parameters, such as color map or clip bounds, are altered.

In order to achieve this, CARTA improves on previous solutions by using a combination of multi-threaded cropping, down-sampling and compressing using the ZFP lossy compression library; a tiled rendering system with multi-tiered data caching; and GPU-accelerated image rendering. Unlike previous radio-astronomy visualization systems, the front-end component is responsible for image rendering, compositing and overlaying of WCS information, leaving the back-end component to focus on data delivery and data analysis, for which full-precision data is required. This has lead to significant performance improvements over the initial CARTA implementation.

We discuss the interface defined between the front-end and back-end components of CARTA, the flow of data from disk to rendered pixels, and future improvements to CARTA that will further improve its efficiency in data delivery.