It may not be self-aware (yet) but this computing monster is ready to take over the world. Well, at least a telescope in Chile.
Say hello to the correlator for the Atacama Large Millimeter/Submilimeter Array, or ALMA. The correlator is the computer that runs at the backend of an array of radio telescopes called an interferometer. It, very basically, combines all the signals of the antennas so that it can function as one single telescope.
The correlator was largely constructed in the building right across from where I did a lot of my graduate work at the National Radio Astronomy Observatory in Charlottesville, Virginia. I used to take any excuse, usually a visiting tour group, to gaze at the supercomputing monstrosity, although they were only working on it in sections. The picture above gives you a better sense of its full size.
Though large (134 million processors) and fast (17 quadrillion operations per second), this computer has just one purpose: to suck in all of the data from ALMA's 66 dishes and transform it into data that can then be sent to the astronomers to calibrate and analyze. The correlator gives the interferometer its power to see incredibly fine detail and small structures, such as protoplanetary disks and distant star-forming galaxies.
The correlator came online in December as yet another step towards completing ALMA, a telescope that will give astronomers an unprecedented look at the sky in millimeter wavelengths. First science results have already been coming in from a partial array and correlator, giving scientists a tantalizing glimpse at what the full power of the array will hold.
Below is a picture of the back of a tiny section of this giant correlator. With 66 antennas, there are over 2000 combinations of antenna pairs, which leads to thousands and thousands of physical connections that must be made between circuit boards by hand. I think it is fair to say that wires were crossed more than once, and a consistent labeling scheme was necessary.