China needs to invest more in supercomputer race

By Zhang Yunquan Source:Global Times Published: 2018/3/19 20:43:39

Illustration: Luo Xuan/GT


This year or next, China may lose its global supercomputer leadership, and how to respond will be a challenge. Wu Lixin, an academician of the Chinese Academy of Sciences and a deputy to the 13th National People's Congress, expressed this concern during this year's two sessions.

It's a reasonable concern. The window for leadership of China-US supercomputing is 2020. With the strong investment of the US in supercomputers, China should establish and start the exascale-class supercomputer system construction as soon as possible to ensure our leading edge in the sector is maintained in the long term.

In recent years, Chinese supercomputers have always been world leaders. However, the US is also making every effort involving supercomputer construction. China's Sunway Taihu Light and the Tianhe, or the Milkyway series are all in the leading class internationally. Supercomputing has become an important symbol of our country's global competitiveness and an international calling card for our technological innovation ability.

However, our pioneering advantages are precarious. In the summer of 2018, the US will finish building two supercomputers with speeds of 200 petaflops and 130 petaflops respectively, at two national labs. The third supercomputer in the US, which is planned to be installed at a third lab, will be a 1,000-petaflop or exascale-class supercomputer. The US is stepping up research and development, and this supercomputer will be finished by 2021.

Very soon, China will release development guides for this sector, and this time two exascale-class supercomputers might be developed. However, in terms of reaching goals for 2020, the overall progress is a bit slow. After all, to ensure our long-term supercomputing advantage, we must keep running.

If advantages in supercomputing are lost, numerous initiatives in various disciplines will be affected. The exascale-class supercomputer is China's next-generation supercomputer, capable of running 1 billion billion calculations per second. It will be a supporting device for all sorts of scientific research.

If we lose the leading position in supercomputing, the country's scientific research will be disadvantaged and opportunities for scientific research involving international cooperation will also be greatly reduced.

As an important support platform for artificial intelligence (AI), the supercomputing ability of a nation helps determine the performance and long-term development of that nation's AI industry.

Further, with regard to data collection and processing, any inferiority in supercomputing capability will delay this process, with a great negative impact on the military and our national security.

With the completion of exascale-class supercomputers, we also need to pay attention to another issue: China's application of supercomputers still lags way behind its development of supercomputing capability. In key areas such as manufacturing and fundamental scientific research, which require large-scale supercomputing applications, most applications are monopolized by foreign countries. We should turn our speed advantage into an application advantage.

Unlike hardware for supercomputers, the development cycle of related software is even longer, perhaps as long as 20 to 30 years. It involves a series of complex processes including problem modeling, the design of algorithms and operation verification, involving a variety of disciplines.

The development of application software has been neglected in China, so our software development is not in line with the development of hardware. The lack of funds, the decentralization of research and development and the shortage of software talent are all problems facing China. To solve these problems, State-level supercomputing parallel application software engineering centers should be established, with multidisciplinary teams and appropriate funding, to formulate medium- and long-term research plans for China's supercomputing applications.

Because scientific research is cyclical, we do not need to be too nervous about 2020, as long as our exascale-class supercomputing system can be successfully established and the level of supercomputing applications can be quickly improved.

In term of winning the "crown" of supercomputing, China is still at an advantage. Amid fierce competition, overcoming weaknesses and being alert will enable us to keep making progress.

The author is a research fellow of the State Key Laboratory of Computer Architecture at the Institute of Computing Technology of the Chinese Academy of Sciences, and director of the National Supercomputing Center in Jinan.


blog comments powered by Disqus