A high-performance, flexible data acquisition library for radio instruments

Abstract

The next generation of radio instruments will generate an unprecedented amount of data, with the Square Kilometre Array (SKA) predicted to generate approximately 160 TB of raw data per second. The computational resources required to process this data are considerable, and alongside this challenge there is an associated one: How is this data going to be transported efficiently? Compute servers composing the processing backends of these instruments must be capable of receiving a high data throughput, with enough leftover resources to perform the required computations. In this paper we present a high-performance and flexible library for packet acquisition. This library was developed as part of the effort to prototype a single Low Frequency Aperture Array (LFAA) station, the low frequency component of the SKA. This resulting LFAA demonstrator, the Aperture Array Verification System (AAVS), generates several types of data streams which must be received and processed. The data acquisition (DAQ) library was designed to be capable of receiving and processing multiple data stream types simultaneously whilst minimizing system load. We present the philosophy behind the design and implementation of this library, as well as provide some benchmarking results.