Research that South African scientists are helping to conduct around rugged microservers and new computing architectures for the SKA (Square Kilometre Array) radio telescope could create powerful, energy-efficient platforms that find wide applications in commercial datacentres around the world.
SKA South Africa, a business unit of the country's National Research Foundation, is working with IBM and Astron, the Netherlands Institute for Radio Astronomy, to research advanced technologies to handle the big data that will be produced by the SKA.
The SKA, to be located in South Africa and Australia, is an international effort to build the world's largest and most sensitive radio telescope with the goal of better understanding the history of the universe. When the SKA is completed in 2024, it will collect large volumes of data from deep space containing information dating back to the Big Bang more than 13 billion years ago.
The microserver research is just one of the seven sub projects that IBM/Astron Dome partnership has launched in their four-year collaboration to develop a fundamental IT roadmap for the SKA. The collaboration is also investigating emerging technologies in nanophotonics, and data streaming.
One key output from the Dome project could give an IBM edge in the emerging market forroughly the size of a bar of soap. This market is expected to boom over the next few years, growing from shipments of around 16,000 units in 2011 to around 1.2 million units in 2016, according to market researcher IHS.
From the project, IBM will be able to evaluate the good and bad of its microserver technology over the next year, and help put it on the road to commercialisation, says Ronald Luijten, the Zurich-based manager leading IBM's Dome project. IBM hopes that the research will demonstrate the suitability of microservers for applications that are data-intensive rather than compute-intensive, such as analytics and big data.
The SKA requires servers that are small, powerful, and highly energy efficient because the designers of the telescope might want to perform a first round of data filtering or analysis close to the antennas, or even on them. These antennas will be located in the desert off the electrical grid. The SKA organisation may also want to use a microserver design within its datacentres. In that case, many of the devices would be packed close together in metal racks.
IBM expects that the microservers it will test for the project will outperform today's servers by a considerable margin on the "figure of merit" – performance per watt per volume, says Luijten. One way that the technology will get better performance from less energy is by packing the components in the server board tightly into liquid-cooled 3D stacked chips.
By reducing the distance between the components and eliminating I/O interfaces, IBM expects to be able to cut power requirements and improve performance. Each microserver board will contain an inexpensive Freescale-built PowerPC processor as well as memory, fulfilling most of the functions of today's larger servers.
One way that IBM is hoping to differentiate its microserver platform from most of the competition is by using hot-water cooling to allow it to stack microservers more densely together, Luijten says. Using water heated to around about 50 degrees Celsius, the chips can be cooled to the desirable temperature of around 85 degrees.
The water is pumped through microscopic channels across the surfaces of chips in a server. The hot water could be reused for another purpose, such as supplying energy for sea water desalination in the southern African and Australian deserts in the case of the SKA. The team in South Africa will look to make the microsevers rugged or "desert-proof" to handle the extreme environmental conditions where the SKA will be located.
Scientists from SKA South Africa will also conduct research into advanced computing algorithms for the capture, processing, and analysis of the SKA data so clear images can be produced for astronomers to study as well as testing and development of sophisticated software analytics.
IBM says the Dome research's implications stretch beyond astronomy, towards an era of cognitive computing where computers can learn and reason. It envisages that the technologies will be as useful in financial services, medical and telecoms datacentres as they are in studying the origins of the universe.