The Computing Group focuses on the measurements that are needed to transform the digital signal of the telescopes into astronomical data. This concerns enormous amounts of data: terabytes per second! Therefore we work with supercomputers, fast glass fiber networks and innovative storage systems. The Computing Group is the largest group in the R&D laboratory.
Mathematics, physics, computing
The Computing Group consists mainly of software engineers. The astronomers are our ‘customers’. We develop algorithms to process data efficiently to astronomical products (such as images of the sky). These algorithms are at in the field of calibration (calculating parameters) and imaging (Fourier transformations). The algorithms must be as efficient as possible. The main challenge is processing the data before new data has been observed.
GPU-supercomputers for correlation
The data of different stations are put together (correlated) on a supercomputer. For one of our telescopes, LOFAR, this happens at the CIT (Centre for Information Technology) in Groningen. Recently, the migration of an IBM Blue Gene P-system to a supercomputer working on GPU’s was completed. Many of the calculations on a correlator are simple, but at the same time a great amount of calculations have to be done. GPU’s are ideal for this work. The Computing Group also examines different accelerators (GPU’s, intel xeon phi) and explores which are most applicable for various applications. This is not only done for LOFAR but also for future radio telescopes such as the SKA and APERTIF. To this end, we focus on performance in terms of flops per Watt.
DOME: the collaboration with IBM
The amount of data we generate is truly astronomical. There are few places where this much data is generated. This also draws the attention of industry. In the DOME project, a collaboration of ASTRON and IBM, we work on innovative algorithms for radio astronomy. We focus, among other things, on how data is distributed and stored efficient.