a great power’s most important weapon | supercomputing: we have become the “number one in the world”
2024-10-02
한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina
"china economic weekly" reporter zheng xue | reporting from beijing
the prosperity of science and technology will make the nation prosper, and strong science and technology will make the country strong.
the development of supercomputers is related to the country's technological future. it is the "crown jewel" in the computer field and an important symbol for measuring a country's technological level.
as an important tool of a great power, supercomputers are mostly used in national high-tech fields and cutting-edge technology research, such as aerospace, astrophysics research, weather forecasting, tidal forecasting, etc. its importance is self-evident.
how fast can a supercomputer perform operations? take the tianhe-1, the first petaflop supercomputer independently developed by my country, as an example. its peak computing speed is 4,700 teraflops/second. if 6 billion people need to calculate a problem in a year, the supercomputer can calculate it in one second. over.
looking back at the development history of supercomputers in china, the “glass room” is a past that cannot be forgotten. in the 1980s, the world's supercomputer technology developed rapidly. western countries have long adopted multilateral export control policies against china. although china has spent a high price to purchase supercomputers from the united states, the related performance has been emasculated. the equipment needs to be placed in an independent computer room. the computer room keys and activation passwords are controlled by the united states, and access to the computer requires review and supervision by us personnel.
key core technologies are difficult to buy, and china needs to have its own supercomputer.
in may 1978, the national defense science and technology commission held a supercomputer program demonstration and collaboration meeting in beijing and determined that the machine was codenamed "project 785". my country officially launched the development of supercomputers.
being poor and destitute was a true portrayal of that time. the technology is backward and the data is scarce, but this cannot stop scientific researchers who are determined to tackle key problems. ci yungui, the "father of supercomputers," even issued a military order: "even if we risk our lives, we must build china's supercomputers."
in december 1983, my country's first exascale supercomputer "yinhe-1" was successfully developed. china became the third country after the united states and japan to independently design and manufacture supercomputers.
taking "yinhe no.1" as the starting point and with the support of many parties, china's supercomputer has embarked on a journey of independent innovation and continuous transcendence step by step.
series of supercomputer products such as galaxy, tianhe, and sunway have gradually gained brilliance, and the computing speed of domestic supercomputers has also achieved breakthroughs from petaflops to exaflops to tens of exaflops.
in 2010, "tianhe-1" won the title of the world's top 500 supercomputers (hereinafter referred to as "top500"); the successful development of "tianhe-2" in 2013 marked that my country has been at the forefront of the world in the field of supercomputers; tianhe's new generation supercomputer, which adopts completely independent innovation, has a computing speed of 10 billion times. in june 2016, my country's first supercomputer "sunway taihulight" built entirely with domestically produced processors ranked first in the top500 rankings with a computing speed that was nearly three times faster than the second place.
"sunway taihulight" is also the world's first supercomputer with a peak computing performance exceeding one billion floating-point operations per second. it has won the gordon bell award, the highest award for high-performance computing applications.
supercomputers have been developed for more than 40 years. with the efforts of generations of scientific researchers, china's supercomputers have achieved a leap from catching up to running parallel to partially leading.
meng xiangfei (second from right) discusses technical solutions with the team
witness: meng xiangfei, party secretary of national supercomputing tianjin center
in 2009, after graduating with a ph.d., i joined the national supercomputing tianjin center (hereinafter referred to as the "tianjin supercomputing center"), the first national supercomputing center under preparation, and participated in the development and deployment of my country's first supercomputer ranked first in the world. "tianhe-1" related work. this experience makes me extremely proud and honored.
later, i participated in the "second entrepreneurship" of supercomputing. after ten years of hard work, the world has truly recognized our achievements in the field of supercomputing.
at present, my country's supercomputing has achieved a leap from catching up to running parallel to partially leading, not only in technology, but also in software and application breakthroughs. relevant achievements are inseparable from the efforts of generations of scientific researchers, and they are inseparable from the inheritance of the ambition, integrity and confidence of scientific researchers.
in 2022, meng xiangfei introduced the development history of china’s supercomputers at a public event.
i didn’t expect that we would become “no. 1 in the world”
in 2006, i was studying for a phd in theoretical physics at nankai university in tianjin. research on related topics required strong computing power, but at that time the country did not have such conditions. later, with the support of the china scholarship council, i went to the united states to study.
i feel a lot about the platform and environment of american innovation capabilities. in the second year after arriving in the united states to study, my laboratory released a total of four research results at the international annual conference in the field of theoretical physics, two and a half of which were mainly completed by me. i felt very fulfilled at the time. but at the same time, i also see another problem: the signatures of american institutions rank first, and our contributions are really the icing on the cake for others.
at the end of my studies, my american tutor wanted me to stay in the united states. when i was confused, on the way to take a walk one day, i remembered the "three patriotic questions" of the old principal of nankai university, and felt that as a chinese, i had not yet contributed to the innovation and development of my country. i decided to return to china. no matter how much or how little i do, it will be ours.
in 2009, i joined the tianjin supercomputing center and participated in the development and center construction of "tianhe-1".
the challenges we faced at that time were not small. we did not choose the mainstream ring architecture route at the time. instead, we used a heterogeneous fusion computing architecture that combined cpu and gpu to build a supercomputer. we were the "first" on this route. .
in fact, relevant research work has begun in 2007. in order to complete relevant tasks, the entire r&d team chose to "retreat" and verified through key underlying technologies, such as software and hardware design, operating system, compiler, software and hardware collaborative scheduling, etc. we are on the right track.
in the summer of 2009, we started the construction of the basic environment and the installation of the complete system. we need to complete the deployment and installation of tens of thousands of optical fibers and tons of cabinets. in a room with a temperature close to 40℃, i would sleep on the floor when i was tired, and then resume work when i woke up. during the equipment installation process, some old experts who were about to retire were also working with us on the front line to pull optical fibers and move cabinets, which gave me an education at that time.
in this way, it took us about 7 months to complete the deployment task that took others more than a year and a half to complete. china's first petaflop supercomputer "tianhe-1" was born.
in november 2010, the ranking of the world's top 500 supercomputers was announced, and "tianhe-1" won the first place.
it was already evening when the news came, and i was on duty with several colleagues. after hearing the news, i was very excited. i never expected that our team would actually become the world's number one project. after my colleagues and i cheered together, we ran a few laps around our "tianhe-1". for me, who is usually calm, it was already very "extraordinary".
"second entrepreneurship" allows china's supercomputing to truly gain a foothold
but soon, questions about whether it was “good enough or not useful” came one after another. international experts questioned that we built a "sports car" just to be fast; some people in china also questioned that we built a large "game console."
the reason is that although supercomputing technology made a breakthrough, it also faced challenges at the time: first, the foundation for supercomputing applications at that time was insufficient, and second, the software level and algorithm model had not yet changed.
for us, what is really difficult is to achieve the “second breakthrough” in applied innovation, that is, to truly transform supercomputing into innovation and productivity.
the "second breakthrough" is in the hands of our young team. as the person in charge of the application r&d team, i was prone to “getting on top” when faced with doubts. i issued a military order: if i can’t implement the application of “tianhe-1”, i will pack up and leave.
we have built an all-round team of pre-sales, after-sales, and r&d. we take the pre-sales team to explore needs and promote it, and are responsible for the promotion of supercomputing systems, capabilities, and innovations. we ride the green train and travel to dozens of cities every year and hold hundreds of lectures or exchange meetings to let everyone know about supercomputers.
the after-sales and r&d teams mainly develop supercomputing technical service capabilities, understand practical and industrial needs, use our supercomputing platform to solve needs, or re-develop operating software or models based on needs to solve technological innovation and industry problems.
it is precisely in the past few years that we have built supercomputers into important tools for great powers that can calculate the sky, the earth, and people.
calculation of days, used in applications such as aerospace flight patterns, weather forecasts, etc. calculation land, used in oil exploration, dam seismic analysis, etc. human computing can be applied to drug research and development and analysis of large-scale genetic data.
computing power is truly transformed into innovation and productivity, secondary innovation is achieved, and various ecologies emerge naturally. focusing on the development of supercomputing, a large number of information technology companies have emerged in the fields of domestic chips, operating systems, servers, etc., laying the foundation for the development of china's information and innovation industry.
the “second entrepreneurship” allowed china’s supercomputing to truly gain a foothold. since 2010, after ten years of hard work, china’s achievements in the field of supercomputing have been truly recognized internationally.
looking at it now, china’s supercomputing development embodies three “qi”:
ambition, the heterogeneous converged computing architecture carried by "tianhe-1" has become the world's first and leads the international supercomputing trend, reflecting our ambition to move forward and compete at the forefront of the field.
backbone, when we first started, we were very poor. our seniors accepted this task and issued a military order, making the challenge impossible.
we are confident that tianhe’s new generation supercomputer, which is fully independently innovative, gives us more confidence in the highly competitive supercomputing field.
of course, supercomputing is also facing new challenges: moore's law is facing the end challenge, and the computing paradigm has changed; artificial intelligence is developing rapidly, requiring large-scale low-precision hybrid computing support, and supercomputing needs to think about how to integrate with intelligent computing; in addition, quantum with the rise of computing, supercomputing will need to develop in tandem with it in the future. now, when we design a new generation of supercomputers, we have already considered these issues in advance.
(this article was published in "china economic weekly" issue 18, 2024)
