Developing super powers
12 January 2018 | 0
In this very space in 2016, I talked about how a new and different arms race was beginning, where supercomputing capacity would be acquired to further, among other things, strategic and defence capabilities for nations. Well, the trends that drove that prediction have come to fruition, with the EuroHPC project which is now expected to be funded to the tune €1 billion, or in US terms, some $1.2 billion.
“The lack of HPC independence threatens privacy, data protection, commercial trade secrets, and ownership of data in particular for sensitive applications”
In the statement on the funding, which also gives some detail of the project, the commission says European scientists and industry increasingly process data outside the EU because their needs are not matched by the computation time or computer performance available in the EU.
“This lack of independence threatens privacy, data protection, commercial trade secrets, and ownership of data in particular for sensitive applications,” said the statement.
That is pretty strong.
When one looks at TOP500, the definitive list of supercomputers in the world, out of the top 10, only one is European. The CSCS Piz Daint in Switzerland is listed as the third most powerful computer in the world, behind two Chinese behemoths, the Sunway TaihuLight and the Tianhe-2. All the same the Piz Daint (named for something Swiss and mountainy) is a Cray X50 (American) with Intel Xeon ES 12C chips and NVIDIA Tesla P100 GPUs—hardly homegrown.
There has long been an initiative in the United States to at least have homegrown alternatives to foreign designed and manufactured chips for critical infrastructure and defence applications.
The Commission statement on EuroHPC funding says the project will develop “European supercomputing technology” including “the first generation of European low-power microprocessor technology, and the co-design of European exascale machines”, all the while fostering applications, skills development and a wider use of HPC. It also says that it will “support the development of exascale (a billion billion or 1018 calculations per second), performance systems based on EU technology, by 2022-2023.”
That is an ambitious undertaking and clearly shows that while the EuroHPC project is happy to look at existing technologies for the first steps, European technologies are preferred for the later, more advanced stages. This is further evidenced by some of the examples cited for the drivers.
HPC is essential for national security and defence, said the commission, and it gives the example of developing encryption technologies, tracking and responding to cyberattacks, deploying efficient forensics and then drops the figurative bombshell of “or in nuclear simulations”. Right so.
That leads to, yet again, rounding off with a quote from Steve Conway, HPC analyst, IDC, who said the topic “is now so strategic that you really don’t want to rely on foreign sources for it”.