South Africa will host one of the fastest supercomputers in the world
The Square Kilometer Array (SKA) project will build two supercomputers that would be the joint sixth fastest in the world if they were finished today, according to Top500.
The supercomputers are being built for the SKA-Mid and SKA-Low arrays. The SKA-Low project is being hosted in Australia, and most of the SKA-Mid array will be in South Africa. One of the computers will be based in Cape Town.
The supercomputers, known as science data processors, will each have a processing speed of around 135 petaFLOPS (quadrillions of floating point operations per second).
Currently, the US-based Frontier supercomputer blows all others out of the water in terms of processing power with more than double the speed of its nearest rival — the supercomputer Fugaku in Japan.
The Frontier supercomputer has a processing speed of 1.1 exaFLOPS (or 1,102 petaFLOPS), which lets it perform over a quintillion (1 followed by 18 zeros) calculations a second.
Morocco’s Toubkal supercomputer, the fastest supercomputer in Africa, appears in 180th place on Top500’s list.
When the SKA’s supercomputers are finished, they will be more than 30 times faster than the Moroccan machine.
The process of awarding the tenders to construct the South African supercomputer will begin in the first quarter of 2024, SKA spokesperson Anim van Wyk said.
It is hoped that construction will be finished within three years from that date.
When completed, the radio telescope will be the largest scientific instrument in the world.
It is hoped that the telescope will be able to unravel secrets of the early universe and test scientific theories such as gravity.
South Africa won the bid for the majority of the project in 2012 as it was selected to host the SKA-Mid telescope, which will be used to detect radio waves between the frequencies of 0.35 gigahertz and 15.4 gigahertz.
Australia won the bid to host the SKA-Low Telescope, which detects lower-frequency radio waves.
The data requirements for the project are immense.
For the SKA-Mid project, the data from all 197 dishes must be collected and sent to the central data processing facility.
The supercomputer will receive up to 8.9 terabits of data per second from the SKA-Mid Central Signal Processor. The data will travel along hundreds of kilometres of fibre-optic cables to reach the science data processor.
Why such a large telescope?
The SKA-Mid telescope comprises many individual dishes rather than a single large one for a few reasons.
Mainly, it becomes practically difficult to construct a large steerable single telescope above a diameter of 100 meters.
The surface area of all the dishes in the SKA-Mid array will be a square kilometre — which would be impossible to manoeuvre as a single dish.
The higher the surface area, the fainter the signal a telescope can detect and the further it can “see” into space.
The other benefit of spreading the telescope out over a wide area is that it allows extracting additional data through a technique called interferometry, helping increase the resolution of measurements.
“Higher resolution is achieved by increasing the separation between the antennas — doing this is the equivalent of creating a zoom lens,” the SKA Observatory (SKAO) explains.
“How? Imagine radio waves of the same wavelength coming from an object in space, and two distant antennas on the ground.
“The waveform is made up of cycles up and down. When it hits antenna number one, it will include part of that up and down cycle; when it hits antenna number two, further away, it will include a different part of the cycle.
“This difference is known as a phase offset, and it tells us about the position of an object in the sky.”
Increasing the distance between antennas increases this resolution.
“If the two antennas were next to each other, there would be no phase offset detectable; in other words, the two antennas would see the same thing,” the SKAO explains.
“The more distant the antennas are, the greater amount of phase offset can be detected, and the more precisely positional information can be recorded.
“This is why aperture synthesis creates higher resolution/a greater amount of detail than is possible compared to collecting radio waves with one dish. These different versions of the waveform are then combined and synchronised to counter the time delay.”
Interferometers like the SKA use a technique called aperture synthesis that enables arrays to act like one big telescope equivalent in size to the distance between the two furthest antennas.
Construction timeline
Van Wyk said that the South African Radio Astronomy Observatory (SARAO) will begin procuring the Cape Town-based SKA-Mid Science Data Processor in the first quarter of next year. It should be completed within three years.
The SKA-Low Science Data Processor in Australia should also be completed in 2027.
There are also regional data centres where data will be stored and become accessible for use that form part of the information network of the project. Van Wyk said it is expected that these will be completed before the end of the decade.
“As for the SKA Regional Centres, the network will be able to support the science verification work starting in 2026, but with capabilities and capacity growing to be ready for full SKA operations by the end of the decade.”
In December last year, R5 billion in contracts were awarded as construction officially kicked off on the project’s next phase.
Some of the dishes are already online, and data is being produced that scientists can use.
Here is an image that was produced of the centre of our own Milky Way galaxy using dishes that are currently operational:
Milky Way galactic centre image surveyed by MeerKAT, shown at its inauguration. Image: SARAO
Loading ...