Russia building 10-petaflop supercomputer
T-Platforms, a Moscow-based tech company that has built some of that nation's largest systems, is developing a 10-petaflop supercomputer for M.V. Lomonosov Moscow State University, the company said this week.
This large system puts it in the ballpark of similarly announced systems being developed in the major supercomputing countries, and may signal Russia's intent to become a major participant in the race to exascale.
Russia is playing catch-up in a rapidly developing race among China, Japan, the U.S. and Europe to build an exascale system in this decade. These are systems which would have 1,000 petaflops of computing power. (A petaflop is a quadrillion floating-point operations per second.)
Building an exascale system will require new approaches in microprocessors, interconnects, memory and storage. If breakthroughs happen outside the U.S., it could seed development of companies that could challenge the U.S. dominance in tech.
Russia "is committed to having exascale computation capabilities by 2018-2020 and is prepared to make the investments to make that happen," said Mike Bernhardt, who writes The Exascale Report and who was asked to respond to a Computerworld query by T-Platforms. More details about the Russian exascale effort will come out in the next year, he said.
T-Platforms is establishing itself as the leading HPC maker in Russia, and is also gaining customers outside of Russia, particularly in Europe. It has previously built a 1.3 petaflop system at Lomonosov.
The latest system at Lomonosov will be water-cooled and is expected to be operational by the end of 2013. It will use Intel and Nvidia chips, and possibly Intel's MIC chip if it is available for design efforts in 2012, the company said.
The view from Russia is very similar to that of Europe. All these nations would like to be less dependent on U.S. technology to build high performance systems.
"At this point, there is unity in believing any company, on a global scale, would be foolish to state that they know the exact technology or components they will use to build an exascale machine," said Bernhardt, in an email. "Systems will be hybrid, heterogeneous and unique. And there are too many unknowns and too many paths being explored at this time to pick the winners from these many options."
"You can expect to see Russia holding its own in the exascale race with little or no dependence on foreign manufacturers," said Bernhardt.
For now, the world is dependent on the U.S. IBM today has nearly 45% share of the systems on the Top 500 list of the world's largest systems; Hewlett-Packard has 28%, and Cray, 5.4%.
Europe is already exploring alternative technologies, including use of ARM processors by the UK-based company ARM Holdings.
"Exascale computing is a challenge, and indeed an opportunity for Europe to become a global HPC leader," said Leonardo Flores Anover, who is the European Commission's project officer for the European Exascale Software Initiative, in an email response to a query. "This can only be achieved if there is a real European policy with the EU member states."
"The EU effort is intended to support excellence in the European supply and use of HPC in all domains (for industry, science and society) that are strategic for us," said Anover. "In particular on the supply side, the goal is to foster the development of a European industrial capability," he said.
China has developed its own interconnects and processors, which it is now using in some of its HPC systems.
There are multiple efforts underway in the U.S. to develop architecture and technologies for exascale platforms. But funding for a multi-year project, which will likely cost billions of dollars is, for now, hinging on a report from the U.S. Department of Energy to Congress by Feb. 10.
This DOE report is expected to outline the reasons for an exascale initiative in the U.S., the international efforts, and the cost of achieving exascale.
(Source: computerworld.com)
