Artist impression of the core of SKA (source: jb.man.ac.uk)
Big science is important, not just because the questions it asks define what it is to be human, but because experiments of such magnitude have the potential to change all the fields of science around them. The Jodrell Bank radio telescope — which after 55 years is still performing valuable science and the world’s third largest steerable dish radio telescope — is an example well known in the UK. More recently, the CERN Large Hadron Collider drove advances in superconducting magnets and in computing, which had to handle 15 petabytes of data per year.
The Square Kilometre Array is the next such project, and its computing requirements are already driving a new round of innovation. The aim is to build a radio telescope with a total collecting area of a square kilometre and spanning 3,000 km from end to end. It will be hosted in the deserts of South Africa and Western Australia, and uses both conventional dishes and aperture arrays. The latter (thousands of them) are steered by real-time computing of the timing of the signals, further increasing the computing demands. When complete in 2025, the sensors will be streaming 10 petabits of data each second and generating 3 exabytes of data each day. For comparison, in 2012 there was around 2 exabytes of data visible on the entire World Wide Web.
With the equipment based in desert environments, most modern infrastructure will not be available; there will be no water to cool servers so computing must be low energy, and the facilities must be reliable enough to require minimal human presence.
Embecosm is a leading provider of not just compiler technology, but low power compiler technology (see James Pallister’s recent blog post). We will be just one of hundreds of companies spread around the world hoping that our technology can play some small part in making this tremendously exciting project a success.