ASKAP Computing

The software systems for ASKAP include scheduling of observations, monitor and control of the telescope during observations, processing of observations into scientifically useful data products, and archiving of data products.

The list of processes to be controlled synchronously within ASKAP is larger than that usually associated with a synthesis telescope, such as the ATCA, where processing and archiving are usually performed asynchronously ("off-line"). This reflects the operational model of ASKAP whereby the immediate result of observing is a scientifically useful data product in the form of images and/or catalogues. A major consideration for ASKAP is that the data rate from the observations will be large, so it is unlikely that more than a few days of 'raw' data will be stored.

The computing load for ASKAP and SKA is shown compared to the performance of the fastest computers in the world.
The computer for SKA will be one of the fastest machines in the world at that time.

The project covers the development of systems for:

Campaign-style observing in the first instance, such as a year-long continuum and HI emission survey of the entire visible sky, together with transient detection. Software to support substantial observing of individual targets is lower priority.
Firmware for the operation of telescope subsystems is to be provided by the relevant ASKAP project.
Development of specific analysis capabilities such as catalogue construction is in scope. However (as noted in the first dot point above), the development of a fully general analysis package for processing ASKAP and KAT observations (based on MIRIAD or AIPS++, for example) is excluded.

The high level architecture will be loose coupling of tightly integrated components.
Each component can be implemented in an appropriate software package.

The major components to be produced are considered to be subsystems, e.g. architecture & design, archive, calibration and imaging methods, central processor, infrastructure, management, telescope monitor & control.

The components will be delivered, deployed, and tested on a sequence of telescope platforms, including the Parkes FPA test antenna (early 2008), BETA (2010), and ASKAP (mid 2012).

The telescope monitor and control software will based on the EPICS open source software, widely used for high energy physics and optical telescopes.

Data processing into scientific data products occurs in the Central Processor - a computing system of architecture appropriate to the data processing needs. Initial analysis indicates that for ASKAP, this will be a relatively large (of order of ten thousand elements) cluster. This will run special purpose code developed specifically for the processing of ASKAP observations.

A schematic diagram showing the data rates and computational loads for the various physical elements of the ASKAP telescope, illustrating how the data flows through the beamformers and the correlator, through the central processor, and eventually into the science archive.

Data products are staged to the science archive as they become available, and are accessed principally via the science archive. The data volumes may be very large (many TB per hour, depending on the details of the telescope such as baseline length), and so it is not yet clear what will be retained in the archive. Initial analysis into catalogues (continuum, spectral, time) will be performed on the Central Processor, but may be repeated later offline if resources are available.

Development of a single digital backend for ASKAP and the SKA

An overview of the single digital backend system for ASKAP and the SKA can be found here (pdf).

The ASKAP systems design is based on FPGA-based beamformers and correlator systems, rather than a single digital backend (SDB) as outlined by the Digital Systems Project as this is a longer-term option (that is, SKA-driven). However, if the SDB R&D proceeds at a faster pace than anticipated we will consider implementing it as a complement to the FPGA systems within ASKAP (for example, to process limited bandwidth as a running test of the SDB).

CSIRO sought expressions of interest from industry via the AusTender website for the development of a prototype single digital backend computing system (CSIRO EOI 2009-002). The tender opened on 11 February 2009 and closed on 23 March 2009.

Opportunities for Industry

Information on opportunities for industry can be found on the industry page.

ASKAP computing and the development of SKA systems

On 10 July 2009, CSIRO hosted a special technical ASKAP/SKA industry discussion session on ASKAP computing challenges and relevance to the SKA. The two presentations from the session are available for download:

Tim Cornwell's presentation (10MB pdf).
Ben Humphreys' presentation (7MB pdf).