An Automated Scalable Framework for Distributing Radio Astronomy Processing Across Clusters and Clouds

TL;DR

This paper introduces a scalable, portable framework for processing large-scale LOFAR radio astronomy data across clusters and clouds, enabling efficient handling of the extensive data from the LoTSS survey.

Contribution

It presents a novel framework that makes LOFAR data processing portable, scalable, and automated, facilitating large-scale radio astronomy data reduction across distributed computing environments.

Findings

Framework successfully processes LoTSS data at scale.

Enables high-bandwidth data access and processing across clusters.

Operational implementation demonstrates practical scalability.

Abstract

The Low Frequency Array (LOFAR) radio telescope is an international aperture synthesis radio telescope used to study the Universe at low frequencies. One of the goals of the LOFAR telescope is to conduct deep wide-field surveys. Here we will discuss a framework for the processing of the LOFAR Two Meter Sky Survey (LoTSS). This survey will produce close to 50 PB of data within five years. These data rates require processing at locations with high-speed access to the archived data. To complete the LoTSS project, the processing software needs to be made portable and moved to clusters with a high bandwidth connection to the data archive. This work presents a framework that makes the LOFAR software portable, and is used to scale out LOFAR data reduction. Previous work was successful in preprocessing LOFAR data on a cluster of isolated nodes. This framework builds upon it and and is currently operational. It is designed to be portable, scalable, automated and general. This paper describes its design and high level operation and the initial results processing LoTSS data.