Supercomputers (Russian market)

Main article: Supercomputers

Russian supercomputers in the ranking of Top500 in the world

Main article: Ranking of supercomputers in the world Top500

Russian supercomputers are regularly included in the ranking of the largest supercomputers in the world. At the same time, Rosatom's computing systems used to model nuclear weapons are never advertised anywhere, although they are most likely the most powerful.

National supercomputer infrastructure of Russia

Main article: National Supercomputing Infrastructure (NSI)

2024: Putin instructed to increase the power of supercomputers 10 times. This requires billions of rubles

Vladimir Putin During the announcement of the message to the Federal Assembly at the end of February 2024, the President of the Russian Federation instructed to increase the capacity of domestic ones by at least 10 times supercomputers by 2030. According to the head, " states this is an absolutely realistic task."

Putin also stressed that the entire infrastructure of the data economy needs to be developed, and asked the government to propose specific measures to support companies and startups that produce equipment for storing and processing data, as well as developing software.

Vladimir Putin

Earlier in 2024, the President of Russia instructed the Cabinet of Ministers to develop and implement a set of measures to increase the computing power of supercomputers in the country by March 1, 2024. In particular, the government is instructed to develop mechanisms for using the archives of state and municipal bodies, library funds to create data sets that can be used free of charge. Responsible appointed the Prime Minister of the Russian Federation Mikhail Mishoustina.

Igor Bederov, head of the investigation department at T.Hunter, market expert at NTI SafeNet, estimated at 2-3 billion rubles the cost of increasing the power of domestic supercomputers by at least 10 times by 2030. According to the expert, it all depends on which ways the task will be achieved.

As for the amount of funds - it is not yet clear here. It is necessary to ensure the creation of our own quantum networks, quantum centers, quantum computers that could exchange data, "Bederov said in a conversation with RIA Novosti.

In his opinion, increasing the performance of computers will have a great impact on the development of artificial intelligence and other software, as well as open up new opportunities for the field of cybersecurity.^[1]

2023

A supercomputer with a capacity of 54 trillion operations per second was launched in Novosibirsk to solve complex problems in physics and biology

At the end of November 2023, a supercomputer was launched at the Institute of Mathematics of the SB RAS in Novosibirsk, which will help scientists solve complex problems in the field of mathematics, physics, biology and other sciences. The performance of the computer is 54.4 teraflops, or 54.4 trillion operations per second. At the same time, 19.4 teraflops will be involved in GPU-nodes, the press service of the university clarifies. Read more here.

Putin instructed to provide scientists with access to supercomputers

On November 24, 2023, President RFVladimir Putin instructed to provide scientists with access to supercomputers. The head of state made the corresponding statement at the plenary meeting of the conference on artificial intelligence technologies. Read more here.

A new supercomputer has been launched at Moscow State University

At the end of August 2023, Moscow State University. M.V. Lomonosov announced the launch of a new supercomputer. Its power will be used, among other things, in the search for new methods of protecting systems based on artificial intelligence (AI) technologies. The performance of the computer is 400 petaflops. Read more here.

Russia has developed a supercomputer platform for data centers at Elbrus

On July 10, 2023, the RSK Group of Companies announced a universal, high-density and energy-efficient cluster solution "RSK Exastrim" for the creation of new generation Russian supercomputers and data centers (data centers). Read more here.

2022: Russia is 10.5 years behind the United States in terms of total supercomputer power

In terms of the total performance of supercomputers (Rmax, real performance), Russia lags behind the United States by 10.5 years, said on November 28, 2022, scientist Sergei Abramov, a specialist in system programming and IT, Doctor of Physics and Mathematics. sciences, corresponding member of the Russian Academy of Sciences. He received such data by analyzing the information contained in the ranking of the top 500 most powerful supercomputers in the world by year, and shared it at an online meeting with the press on November 28. At the best time - in 2009-2010. - Russia's lag behind the United States in this parameter was 5.5 years.

One of the points that the analysis of the top 500 data shows is that in 2001 both Russia and China in terms of the available real performance of supercomputers were in the same positions. Then China rushed up, caught up with the United States, and from 2019 it seemed to stop. This coincided with the moment when the United States began to impose sanctions against leading Chinese supercomputing companies, prohibiting the transfer of technology to them, provide technical support, etc., Sergey Abramov explained: "This was done simply from the point of view of economic competition. There were no other prerequisites. "

After that, China decided not to publicize its top projects in the field of supercomputers. Thus, a large segment of the industry - leadership systems - fell out of the top 500. China

There are non-public systems in other countries, including Russia, says Sergei Abramov. But in China, the "underwater" part is much larger than in other countries. In China, according to information circulating in the industry, there are two systems of exaflops performance level. And in Russia, the "invisible" part of supercomputers is significantly smaller. It is known about at least two petaflops-level systems under two Russian state corporations that do not "glow" in public ratings.

The leading countries in the field of high performance computing (HPC) are characterized by the presence of national programs in the field of high performance computing, developed state supercomputing infrastructure. Government-funded programs assume that powerful high-performance systems are one of the critical factors in ensuring the security and competitiveness of the economy, since they significantly speed up the solution of many tasks.

The US was among the first to adopt the idea. They have been investing billions of dollars in the development of supercomputing technologies for years, and even with not scarce funding, market players are calling on the American government to invest even more.

In a recent study in the field of HPC, sponsored by Dell and AMD, there is even a reference to the American constitution as one of the arguments for increasing government funding. In accordance with the constitution, the state's task is to ensure protection and universal well-being, and supercomputers help in this: "Nations that do not invest enough in infrastructure and personnel in the field of high-performance computing seriously risk weakening national protection and economic lag."^[2].

In 2022, the United States made a big leap, increasing the gap between itself and other countries in terms of available computing power: it launched the Frontier supercomputer with a real performance of more than 1 exaflops, becoming the first country to acquire a system with such power. Frontier was built by HPE and installed at the Department of Energy's (DOE) Oak Ridge National Laboratory (ORNL). The project was announced back in 2018, it was estimated at $1.8 billion.

Despite the fact that Frontier at first proved to be very "buggy," the beginning of a new era of performance is laid^[3]

In the November list of the top 500 supercomputers in the world, Frontier ranks first. At the time of launch, it accounted for a quarter of the total power of all systems in this rating. 480 out of 500 systems in the November version of the top 500 rating are almost 50 times weaker than Frontier, Sergei Abramov cited data. The stratification between countries in this regard is increasing, he notes: the strong become stronger, the weak become weaker.

The November rating of the top 500 includes 7 Russian supercomputers. With the exception of the system installed at Moscow State University, all of them are used by business: three Russian supercomputers from the top 500 supercomputers are installed at Yandex, two at Sberbank and one at MTS. They are used to solve corporate problems.

In Russia, there is virtually no state supercomputer infrastructure, Sergei Abramov believes. The country has previously had directed programs in this area. So, for example, in 2009-2010. completed the programs "SKIF" and "SKIF GRID" for the development of supercomputers, and in 2010 the list of top 500 included 11 Russian high-performance systems. Among them, many were their own developments.

Good indicators of those years, of course, cannot be fully attributed to the merits of state programs - their funding was small, the scientist notes. At the same time, such programs created a drive in the market, the players tried to develop, attracting, among other things, extra-budgetary funds.

The state and then allocated funds for various supercomputer projects, but some visible results are difficult to allocate, Sergey Abramov believes. In this area, it is not only about money, it is also important to competent organization of work, state cooperation, science and business, well-thought-out management and financing structures. This requires a national program, including, among other things, the regulatory part - how what is created under the program will work in the country's economy.

President Vladimir Putin previously set the government the task of submitting proposals to increase the capacity of Russian supercomputer centers, which should then be included in the national project "Science" and the national program "Digital Economy." According to the idea, by 2031, an extensive system of supercomputing centers should reach design capacity in Russia, providing wide access for institutions of Russian science and higher education to the most modern infrastructure of high-performance computing and supercomputer modeling.

By the end of 2021, as part of this initiative, it was planned to revise the concept of the development of supercomputing infrastructure in accordance with the new goals, to determine the model and structure of the network of supercomputing centers, stages and parameters of productivity growth for ten years to come. For this, a special interdepartmental working group was created in the field of high-performance computing^[4]. There has been no public news about the results of this work since then.

In Russia, at the moment, there is no thought-out system program for the creation and development of supercomputer infrastructure, says Sergey Abramov.

Meanwhile, he says, for the third year, work is underway to write and start a program to create a unified supercomputer infrastructure for all its member countries within the framework of the Eurasian Economic Community (EurAsEC). The concept has already been written and approved at the level of the governments of the participating countries, but the program has not yet started here.

2021

The Ministry of Industry and Trade allocated ₽7,6 billion to create a supercomputer with Russian architecture

The Ministry of Industry and Trade allocated ₽7,6 billion to create a supercomputer with the NeuroMatrix processor architecture, developed by JSC NTC Module. The corresponding tender became known on November 9, 2021, its winner will be chosen in about a month. Read more here.

The government is revising the concept for the development of supercomputer infrastructure in Russia

The Russian government is revising the concept for the development of supercomputer infrastructure. This is reported by TASS with reference to the press service of the Ministry of Science and Higher Education of the Russian Federation.

It is noted that the concept of the development of supercomputer infrastructure will be revised by the end of 2021 in accordance with the new goals. It is also planned to determine the model and structure of the supercomputer center network, stages and parameters of productivity growth for ten years to come.

To solve this problem, an interdepartmental working group in the field of high-performance computing was created under the Council of the Ministry of Education and Science of the Russian Federation for Digital Development and Information Technologies under the leadership of Deputy Minister Alexander Narukavnikov. It includes representatives of the largest supercomputer centers and specialized scientific organizations of the country, as well as the National Research Computer Network (NICS).

According to Alexander Matyunin, head of the applied computing infrastructure department of the digital development department of the Ministry of Education and Science of Russia, by mid-November 2021, the working group will conduct an expert assessment of what kind of supercomputers and high-performance computing will be in demand in key areas of Russian science, education and industry.

The development of a new concept for the development of Russia's supercomputer infrastructure is carried out as part of the execution of the instructions of Russian President Vladimir Putin. He demanded that the Government of the Russian Federation work out proposals to increase the capacity of domestic supercomputer centers, in order to then include them in the national project "Science" and the national program "Digital Economy."

In mid-October Russian Academy of Sciences , president Alexander Sergeyev said it Russia was more than 100 times behind the world-leading performance. supercomputers^[5]

Release of a podcast with the head of the IPS RAS Sergey Abramov on the current state of the supercomputing industry in the Russian Federation

2020

Head of IPS RAS Sergey Abramov: Russia is 12.5 years behind the United States in supercomputing power

If we calculate the lag in years from other countries in terms of available real computing power, then Russia is 12.5 years behind the United States and 9.5 years behind China. Such data at the end of November 2020 were cited by one of the domestic gurus in the field of supercomputers, head of the Institute of Software Systems of the Russian Academy of Sciences (IPS RAS), Doctor of Medical Sciences Sergei Abramov, having analyzed the latest edition of the rating of the top 500 largest supercomputers in the world. Although there was a time - a "nostalgic 2010" - when even America we were only 5.5 years behind, he added.

In 2020, almost all countries present in the list of the top 500 most powerful computing systems in the world increased their supercomputing power. A special breakthrough was made by Japan, whose supercomputer took first place in the latest edition of the list. And Russia, according to the calculations of Sergei Abramov, over the year increased the lag behind other countries by exactly one year.

But I was told the right term was "delayed development," not lagging. Because these 12.5 years mean the following. If America stops and stands still for 12.5 years, and at the same time we repeat exactly the same rate of development that was in the previous 12.5 years in America, then we will catch up with them. But they will not stop, - said the acting director of the IPS RAS.

The academician also cited data that in 2010 Russia held in its hands about 2.5% of the world "pie" of supercomputer power, and at the end of 2020 the same indicator is 0.38%.

The head of the IPS RAS also drew attention to which processors Fugaku is built on, the current most powerful supercomputer in the world, created in Japan - on ARM processors. Sergei Abramov recalled that these processors were previously talked about a lot, the Russian processor manufacturer Baikal Electronics, in particular.

By the end of 2020, all leading countries have built a normal, hierarchical, balanced supercomputer cyber infrastructure, and it is absent in Russia, said Sergei Abramov. And in 2019, the supercomputer community said that Russia needs to re-create a national supercomputer infrastructure.

We also talked about the circumstances that are necessary for this, but are absent, and on the other hand, about those circumstances that are harmful and noticeable, "Abramov emphasized.

Among the first, he attributed the need for an adequate assessment of the state of affairs, a balanced state program and a reliance on all the resources of the industry. And to the second - the "nightmare" of business, the fall in the activity of consumers of supercomputing technologies, the fragmentation of the resources of the industry.

He also recalled the influence on the industry of the arrest of the general director of T-Platforms Vsevolod Opanasenko in 2019, who was accused of fraud. It was a company that made ARM servers suitable for supercomputers. And Russia could, if not repeat the "Japanese miracle," then at least take steps in this direction, Sergei Abramov believes. In 2019, they even sent an appeal to President Vladimir Putin about Opanasenko, and, says Sergei Abramov, there were answers from different departments. But "it is clear that the reaction is not adequate to the severity of the problem," concluded the head of the IPS RAS.

In terms of the number of systems in the top 500 in 2020, China and the United States are leading, between which a serious supercomputer race has unfolded: of these countries, 212 and 113 supercomputers are present in the November ranking, respectively. Moreover, the US Department of Energy has repeatedly stated that their country must definitely win this race: "whoever leads the world in the field of high-performance computing will have a huge competitive advantage in all sectors of the economy."

Rosatom marks the only supplier of software for supercomputers for government agencies

On August 31, 2020, it became known about Rosatom's intention to make its subsidiary, the Russian Federal Nuclear Center, the All-Russian Scientific Research Institute of Experimental Physics (RFNC - VNIIEF), the only supplier of supercomputer modeling systems for departments and state-owned companies for 2020-2024. The state corporation sent a letter with a corresponding proposal to Nikolai Arkhipov, director of the government's defense industry department. Read more here.

Putin decided to join the "supercomputer arms race"

Vladimir Putin instructed to increase in Russia the capacity of computing resources of supercomputer centers, including regional ones, for high-performance computing of scientific and educational organizations. The corresponding order of the president was published on the website of his administration on April 10, 2020.

The instruction of the head of state also provides for the establishment of a procedure for the interaction of supercomputer centers with each other, with scientific and educational organizations on the basis of a new generation national research computer network, an increase in its bandwidth and territorial accessibility.

Measures to achieve the indicated goals should include the national project "Science" and the national program "Digital Economy." The issue of their financial support should also be worked out. Prime Minister Mikhail Mishustin has been appointed responsible for the execution of the order, the term of which is March 1, 2021.

The initiator of the order is First Deputy Director of the Institute of Computing Technologies of the Siberian Branch of RANNDREI Yurchenko. When interacting with the government, he will monitor the execution of the order, the Siberian branch of the Russian Academy of Sciences said.

During a meeting of the president with representatives of the public, held in Cherepovets on February 4, 2020, Andrei Yurchenko noted that for Russia to become one of the five world scientific powers, it is necessary to develop a supercomputer infrastructure in it. At the same time, he drew the president's attention to the facts that indicate a critical lag in the regions of Russia in the development of the supercomputer industry.

Especially noticeable, according to the scientist, this lag becomes in the distance from the center of Russia. Thus, the Siberian branch of the Russian Academy of Sciences, which includes 144 research institutes and federal research centers (FIC), 170 universities and unites over 11 thousand scientists, has less than 1.5% of the country's supercomputer resources. This is critical for ensuring the required indicators for the development of scientific and innovative activities in the regions.

In Yurchenko's opinion, "it will be right to develop a network of sufficiently powerful supercomputer centers that they are located not only in, To Moscow but also in, in St. Petersburg, in Novosibirsk, in Kazan." Vladivostok

Of course, centers are needed not only super-powerful, but also of the second level, as we call them. They should also be in the regions: Tomsk, Tyumen needs such centers, Khabarovsk, Yekaterinburg. Such centers can become a base, the basis for creating, among other things, networks of "data factories," - said Andrei Yurchenko at a meeting with the president.

The Siberian branch of the Russian Academy of Sciences draws attention to the fact that the supercomputer industry is one of the most competitive in the world.

A real "arms race" has unfolded in it, it provides owners of high-performance computer systems and end users with great opportunities in the field of scientific and technological leadership. Russia's entry into the top five leading scientific and technical powers in the world is impossible without ensuring access to research and educational institutions to such systems, and first of all - in strong non-capital scientific and technological centers, since there is objectively the highest unsecured need in the country, the Siberian branch of the Russian Academy of Sciences says.

Particular attention was paid to the development of supercomputers in Russia during the reign of Dmitry Medvedev. So, in July 2009, at a meeting with members of the Security Council, they were tasked with equalizing Russia with the West for the production of supercomputers. Following this meeting, a draft document "Fundamentals of State Policy in the Field of Creation and Application of Supercomputing and Grid Technologies" was adopted, according to which the production of supercomputers in Russia was to be included in the long-term target program "Information Society (2011-2018)." At a meeting of the commission on the modernization of the economy in July 2009, Medvedev announced that 2.5 billion rubles would be allocated from the federal budget for the development of new generation computers.^[1]

However, in the next few years, no active state activity related to supercomputers was observed. They returned to them again only in 2016 - on August 7, an order was issued from the resident to increase the efficiency of the development of supercomputer and grid technologies.

Meanwhile, the gap in the volume of computing power between Russia and the leading countries continues to widen. Thus, the updated ranking of the most powerful supercomputers in the world, published in November 2019, contains only three Russian supercomputers: Christofari of Sberbank in 29th place, Lomonosov-2 of Moscow State University in 107th place and SuperComputer of the main computing center of Roshydromet in 465th place. Then, as in June 2019, the last two were in higher positions - in 93rd and 364th places, respectively.

2018: Russian supercomputer becomes a mass phenomenon

According to the latest 28 editions (spring 2018) of the list of the Top 50 most powerful computers in the CIS, presented RCC of the Lomonosov Moscow State University and JSCC RAS at the International Scientific Conference "Parallel Computing Technologies (PaVT) 2018," there has been a noticeable increase in the performance of the systems included in the list^[2].

The total performance of systems from the Top 50 list on the Linpack test for six months increased from 8.7 PFLOPS (quadrillion (1015) floating point operations per second) to 10.7 PFLOPS. The total peak performance of the list systems at the same time reached 17.4 PFLOPS, although six months ago in the previous version of the list it was only 13.4 PFLOPS.

In just six months since the publication of the previous Top 50 rating, the list has been replenished with nine completely new supercomputers, and seven systems from the list have undergone a large-scale update. These figures, according to the compilers of the Top 50, are a record over the past six years.

There were more applications for entering the rating, but not all systems overcame the lower threshold of performance: to enter the current edition of the Top 50, performance was required on the Linpack test at least 42.6 TFLOPS against 38.1 TFLOPS in the previous edition.

The compilers of the rating note that the main changes in the latest edition of the Top 50 fell not on the leaders of the list, but on systems with a performance of 50-70 TFLOPS. This suggests that the supercomputer in Russia has become a fairly massive phenomenon, and its performance is laid not for the sake of records, but already taking into account the needs and financial capabilities of the customer.

Russian science and education remained the main consumers of supercomputing capacities, they, as in the previous edition, accounted for 18 systems from the rating. The number of systems for specific applied research decreased from 16 to 14. The number of HPC systems in industry also declined slightly, from HPC to four; the number of systems in the financial area remained equal to three.

By the number of systems included in the list, Hewlett-Packard Enterprise remained the leader - 13 systems, as in the previous edition. It is followed by a group of RSK companies with 12 systems, which added another supercomputer to the rating in six months. The top three is closed by the T-Platforms company with 11 systems against seven systems in the previous edition. In fourth place is IBM with five systems (seven in the last ranking). For the first time, four systems appeared on the list, the manufacturer of which is Nvidia.

2015: Accounts Chamber: the state spent 4.9 billion rubles. on non-working supercomputers and grid networks

On November 13, 2015, it became known about the conclusions of the Accounts Chamber regarding state spending on the creation of supercomputers and grid networks for strategically important industries - the lack of practical results^[3].

The state spent 4.9 billion rubles. on the creation of supercomputers and grid networks, but after the abolition of the Presidential Commission for Modernization, they forgot about this project, the Accounts Chamber established.

The Accounts Chamber, during an audit of the use of budget funds aimed at financing the creation of Innopolis, found the lack of practical results in the budget project - "Development of supercomputers and grid technologies." The project was approved in 2009 by the commission under the President [Dmitry Anatolyevich Medvedev 'Dmitry Medvedev]] on the modernization and technical development of the economy.

The goal of the project is to create domestic supercomputer simulation technologies for enterprises of high-tech industries.

"Historically, to work with supercomputers, it was necessary to have physical access to them," explains Ilya Massukh, head of the Information Democracy Fund, former deputy minister of communications. - With the advent of the Internet, it became possible to gain remote access to supercomputers. But conventional communication channels are not enough to transmit all the results that supercomputers receive, and therefore it became necessary to create a dedicated communication network. "

During 2010-2012, 4.93 billion rubles were allocated from the federal budget for the implementation of the project, including 750 million rubles. - through the Ministry of Communications (the ministry submitted these data to the analytical center under the government in 2013). The main consumer of budget funds is the state corporation Rosatom (in 2010 it received 1.1 billion rubles for these purposes).

In 2012, the construction of the grid infrastructure was completed. According to Rosatom, by this time the total performance of super-computers installed in the computing centers of the state corporation amounted to 2610 Tflops with 1400 Tflops included in the project. Within the framework of the project, they created a basic series of compact super-computers, which made it possible to reduce the thematic lag from the United States by 50 times. We created software for imitation modeling and virtual models for the aircraft industry, the nuclear industry, the rocket and space industry and the automotive industry.

The commission was abolished - super-computers remained

The Accounts Chamber noted: the introduction of supercomputer and grid technologies by project consumers (strategic industries) did not happen. 160 million rubles, which in 2012 the Ministry of Communications was to receive from the federal budget for the project, the ministry transferred Innopolis for construction in Tatarstan (in that year, Nikolai Nikiforov, an active supporter of Innopolis development, became the Minister of Communications). Grid networks have not received further funding.

No practical results have been identified in the development of supercomputing technologies

The reasons for the loss of interest in the super-computer project and grid network are that in the summer of 2012, after Vladimir Putin became president, the Modernization Commission was liquidated. Instead, the Presidential Council for Economic Modernization and Innovative Development was formed. As the Ministry of Telecom and Mass Communications reported to the Accounts Chamber, the issue of continuity by the Council of the commission's developments, and, accordingly, the issue of the procedure for making changes to projects previously approved by the commission, was not resolved.

Security Council is looking for a responsible

In the spring of 2013, the Ministry of Telecom and Mass Communications made a report on the construction of a grid network at a meeting of the interdepartmental commission of the Security Council on information security. According to the Ministry of Telecom and Mass Communications, which is given in the document of the Accounts Chamber, for the further implementation of the project and the use of its results, it is necessary to determine the state customer - the project coordinator.

After hearing the report of the ministry, the commission recommended that the Government pay attention to measures for interdepartmental coordination, including the determination of the authorized body, as well as a set of measures and funding for further periods. At the moment, the President is expected to work out the expediency of further measures to develop grid networks in the apparatus of the Security Council.

Grid-network

As of November 2015, there are 21 facilities in the grid network on the territory of nine Russian cities. Including, seven objects were installed in Moscow: in the supercomputer center of the Russian Academy of Sciences, at Moscow State University (now its equipment has been transported to the server department of the Ministry of Communications), at the Voskhod Research Institute, as well as at communication centers on Varshavskoye Shosse and Butlerova Street.

There are two grid network objects in the Moscow region: at the Institute of the Russian Academy of Sciences in Chernogolovka and at the Joint Institute for Nuclear Research in Dubna. There are three facilities in St. Petersburg: at St. Petersburg State University, St. Petersburg State Polytechnic University and at the РТКомм.ру site (a subsidiary of Rostelecom).

There are facilities Nizhny Novgorod in (at Nizhny Novgorod University), (Rostov-on-Don at Southern Federal University), Yekaterinburg (at the Ural Branch of the Russian Academy of Sciences) Kazan and in (at the Kazan Center of the Russian Academy of Sciences). In each of these four cities, one more object is installed at the РТКомм.ру sites.

Also, one object is located in the Nizhny Novgorod region - in Sarov (at the All-Russian Research Institute of Experimental Physics).

In 2012, work on the project was carried out by the Voskhod Research Institute, with which the Ministry of Telecom and Mass Communications signed a contract in the amount of 180 million rubles. It was signed according to the results of the auction with a starting price of 330 million rubles. With these funds, design documentation and organizational and administrative documents were created for the commissioning of the Russian grid.

As of November 2015, the Ministry of Communications annually conducts an inventory of the grid system. As of the end of 2013, it operated equipment with a total value of 260 million rubles. This is the technology of the network control and monitoring center, DSS, switching cabinets, routers, etc.

The Ministry of Communications did not comment on the conclusions of the Accounts Chamber on this issue. Ilya Massukh believes that the grid project was very necessary, and the reasons for stopping it should be sought in the political plane.

2014

Year results and prospects

2014 leaves the impression of a rather unpredictable and ambiguous year in the history of the development of high-performance computing technologies in Russia and the world. Despite the world records set by domestic manufacturers of supercomputing systems, the positive trend of improving Russia's position in TOP-500, which developed in the country by the end of 2014, the general political and economic situation promises not the best times in the medium term^[4].

Path to Exaflops Computation

The Tianhe-2 supercomputer system (Milky Way 2) of the Chinese National University of Defense Technology has led the TOP-500 for the fourth time in a row with a performance of 33.86 Pflops in the Linpack test - it has not changed over the past two years. The configurations of the Titan Cray XK7 (17.59 Pflops) and Sequoia (17.17 Pflops) systems, which are almost doubled and occupy the second and third places, respectively, have not changed. The only addition to the top ten supercomputers in the latest release of the TOP-500 is the shortlisting Cray CS-Storm system of an unnamed US government department with a capacity of 3.57 Pflops.

Starting from the fifth position of the rating, the power of systems is already measured by petaflops units, and starting from the 51st position - by hundreds of teraflops. Progress is noticeable here: in the previous 43rd edition TOP-500 six months ago, there were 37 systems with a performance of more than 1 Pflops. At the bottom of the rating, productivity gains were minimal over the past two decades, although the system, which occupies the last 500th place in the new version of the rating, was in 421st position six months ago.

A curious cut is the graphics accelerators and coprocessor modules used in modern supercomputers, which have a significant impact on the total performance of systems in certain tasks. So, the first and seventh systems of the leading ten use Intel Xeon Phi coprocessors, while the second and sixth have Nvidia GPUs. In total, 75 of the systems included in the last list of TOP-500 operate using accelerators and coprocessors (a year earlier there were only 62 of them). Fifty of them use Nvidia graphics chips, three work with ATI Radeon GPUs, 25 systems are made using Intel MIC (Xeon Phi) technology.

Intel processor solutions, as before, are the platform for the vast majority of systems in the TOP-500 rating (85.8%). The rest of the platforms are represented (in descending order) by IBM Power, Fujitsu SPARC64 and AMD Opteron processors.

Countries and continents

Despite the fact that the most powerful supercomputer on the planet is still located in China, the United States remains the country with the largest presence in the TOP-500 (231 systems), but somewhat gives up its position (in November 2013, the United States was represented by 265 systems). The total number of supercomputers from Asia presented in the ranking also decreased - from 132 to 120 (specifically Chinese - from 76 to 61). The number of European systems increased from 116 in June 2014 to 130 in November 2014.

In the summer, 43rd ranking TOP-500 the presence of Russian systems decreased to the minimum five systems in recent years, while the best of them, the supercomputer "Lomonosov" at, MSU occupied the 42nd line of the world ranking. In the latest, November edition of the rating, the situation has changed: there are nine Russian supercomputers in the composition. The new computing cluster based on MSU the Xeon processor platform, Nvidia K40 accelerators and Infiniband FDR interconnections, created by "," T-Platforms came in 22nd place TOP-500 with a performance of 1.849 Pflops.

It first appeared in the ranking and immediately successfully "landed" in 189th place with an indicator of 289.5 Tflops system of the Russian company Niagara Computers, made on the platform of Supermicro solutions, Intel Xeon processors, Nvidia K20 accelerators and Infiniband FDR interconnections.

The impressive results of the year were demonstrated by the Russian group of companies RSK. The number of supercomputers of its production in the ranking has TOP-500 doubled: the new edition includes four RSK systems with direct liquid cooling, including those developed for St. Petersburg State Polytechnic University (SPbPU) cluster system "Polytechnic RSK Tornado" based on 14-core Intel Xeon processors (658 Tflops, 81st place) and a supercomputer based on a massively parallel PetaStream RSC system with Intel Xeon processors and 60-core Intel Xeon Phi 5120D coprocessors (170.5 Tflops, 390th place). As a result, after commissioning, which is scheduled for next year, the total peak performance of the new supercomputer center SPbPU will exceed 1.1 Pflops.

At the 133rd position of the list, built RSKsupercomputer MVS-10P JSCC RAS (523 Tflops), and 190th place is occupied by a computing cluster (473 Tflops) (South Ural State University SUSU) in Chelyabinsk.

In the overall standings, the leaders in the number of supercomputers included in the TOP-500 are: HP (179 systems, two of them in Russia) and (IBM 153 systems), the indicators of both companies have decreased since the previous rating. Closes the top three Cray with 62 systems.

Transformation

The HPC market is in a period of significant change. According to Rajib Hazra, vice president of Intel and head of the company's technical computing group, the current transformation of the supercomputer market should not be described in the words "better" or "more," rather it is a fundamental change in the use of technology through integration, joint development, improvement of software platforms and even changes in business models of providing HPC computing, primarily in the form of supercomputers as a service.

The further development of the HPC market at Intel is associated with a deeper integration of various system components on a single chip as part of the development of the Intel MIC (Many Integrated Core) architecture. At SC "14, the company introduced the new, third generation Intel Xeon Phi processors, codenamed Knights Hill, which will be launched using a 10nm process technology. Knights Hill chips will be introduced after the Knights Landing generation, the first systems based on them are expected in 2015.

Knights Landing is expected to support more than fifty leading companies in the industry, with many systems using a PCIe expansion board module. Knights Landing chips, in particular, will be used in the Trinity supercomputer of the joint project of Los Alamos and Sandia National Laboratories, as well as in the Cori supercomputer of the National Scientific Computing Center for Energy Research of the US Department of Energy.

With the development of Omni-Path architecture, Intel said it will be able to transmit data at 100 Gbps and 56 percent reduction in switching latencies in midsize and large clusters compared to alternative InfiniBand-based solutions. Through a 48-port switching chip, this architecture will help improve port density and system scalability.

As part of this project, Intel launched the Intel Fabric Builders Program to form a collaborative ecosystem of solutions based on Intel Omni-Path Architecture.

Nvidia has unveiled the new flagship Tesla K80 dual-processor graphics accelerator, which has almost twice the performance and twice the memory bandwidth of its predecessor. Tesla K40 The Tesla K80 accelerator has 4992 parallel CUDA cores, 24 GB of GDDR5 memory and up to 480 GB/s bandwidth. The novelty supports Nvidia GPU Boost dynamic frequency change technology and provides performance up to 8.74 Tflops for single-precision computing and up to 2.91 Tflops - with double accuracy.

Sumit Gupta, general manager of Nvidia (Nvidia) and head of Tesla Accelerated Computing, speaking about the features of Tesla K80 stressed that the new dual-processor accelerator is equipped with a passive heat sink and therefore is positioned as a solution for server systems with centralized cooling. The maximum performance gain when using Tesla K80 accelerators can already be obtained in more than 280 scientific, engineering, commercial and corporate applications.

It also became known about plans to introduce a high-speed NVLink interface designed for direct communication of several Nvidia GPUs in one system, while the processor platform can be any choice - IBM Power, x86 or ARM. The NVLink interface offloads the data bus between the GPU and the CPU and allows you to speed up data exchange by 5-12 times compared to today's systems, which in general can provide an acceleration of calculations by 50-100 times compared to today's record holders.

The introduction of the NVLink bus will begin in 2016, with the introduction of the next generation Nvidia GPU architecture with the working name Volta. Such accelerators with the NVLink topology are supposed to be used in future Summit and Sierra supercomputers, the installation of which is scheduled for 2017.

By then, the NVLink interface is expected to be fully operational on systems with any IBM Power, x86 and ARM architecture processors. In addition, the Volta architecture and NVLink technology will allow you to come close to creating systems of the so-called "exaflops level."

Among the promising developments is IBM's new "data-centrist" approach. He expressed himself in the open architecture of OpenPOWER. The company did not present anything special for the SC "14 conference, but the $325 million contract with the US Department of Energy to create supercomputers for the Livermore and Oak Ridge laboratories speaks of the success of the OpenPOWER Foundation, which includes Google, Mellanox, Nvidia (Nvidia), Tyan, Samsung, Hitachi, ZTE, QLogic, Rackspace and others.

Fujitsu recently unveiled plans for new SPARC64 IXfx processors focused on producing supercomputers with the potential to take the 100 Pflops barrier. The 32-core SPARC64 IXfx processor with two "assist" cores and Tofu inter-unit bus support is expected to provide performance of the order of 1.1 Tflops, which is about 3.2 times the capabilities of the current Sparc64 IXfx processor in double-precision floating point calculations and 6.1 times more in single-precision calculations.

Fujitsu plans to bring the development of the K supercomputer architecture "to the exascale horizon," which, according to the Japanese Ministry of Education, Culture, Sports, Sciences and Technology, will be achieved in 2020.

The emergence of new processor players on the HPC market, such as ARM or MIPS with their 64-bit processors and large-scale commercial programs with their participation, is early to be expected - there is a need to refine the software ecosystem.

Near future

Analysts do not suggest the emergence of an "exaflop" system in the near future. However, they expect serious changes in the top lines of the TOP-500 in 2015. The source of these changes will be systems from China and Japan.

Cray's plans to release Trinity and Cori supercomputers for U.S. science labs represent a major bid. As of December 25, 2014, it is known that both systems will be implemented on the Intel Xeon Phi coprocessor platform with redesigned Silvermont (Knights Landing) architecture, using the new generation of Intel Xeon processors, DDR4 DRAM RAM and Cray Aries high-speed inter-block topology.

In 2017, two supercomputers based on IBM Power9 processors and Nvidia Tesla graphics accelerators with an NVLink interface, created by order of the US Department of Energy, can claim leadership in the ranking. The peak performance of the Summit system for scientific tasks, which will work at the Oakridge National Laboratory, will be 150-300 Pflops. The Sierra supercomputer with a peak performance of more than 100 Pflops will become the basis of the computing system at the E. Lawrence Livermore National Laboratory.

Analysts noted that the world's leading companies in the field of HPC have recently begun to seriously use the capabilities of liquid cooling of cluster systems and the development of the Russian company RSK have outstripped many world brands by years, since the company initially practices cooling working blocks with "hot water."

The RSC PetaStream supercomputers implement a power subsystem developed jointly with Emerson Electric based on the industry standard of direct current power supply with a voltage of 400 V. This allows you to reduce the cross section of the supply buses, increase the efficiency of electricity distribution by more than 90%, increase reliability and energy efficiency while reducing operating costs. In addition, the PetaStream architecture, according to its creators, scales well and can be used to simulate the loads of the very "exascale horizon," including the development of relevant applications.

New, most powerful supercomputer in Russia at Moscow State University - 2.57 Pflops

On November 18, 2014, Russian supercomputer manufacturer T-Platforms announced that it was completing testing and commissioning of a new high-performance computing system for Moscow State University with a peak performance of 2.57 Pflops. In the November ranking of the Top 500 most powerful supercomputers in the world, this system took 22nd place. Details.

2012

Super Computer Financing Scheme

Costs of creating a super computer

GAP-analysis of the RAS of the development of the situation in the supercomputer industry until 2020

Negative outlook for the industry

Five years later, due to bureaucratic obstacles, the technology platform of supercomputer computing did not work. The main oil producers continue to carry out 3D- and 4D modeling of fields on their own. Due to high economic risks, supercomputing and grid technologies are not introduced in heavy and light industries. The state finances this direction at the rate of 700 million - 1 billion rubles. per year. The main costs of the state go to the military-industrial complex, in particular, to the calculations of nuclear tests.

Positive outlook

In the next five years, the project exaflops card was fully adopted along with the creation of a domestic element base: processors, memory cards, data transfer and storage nodes. In total, the state spent 65 billion rubles to create such a supercomputer. This machine entered the top 3 most powerful supercomputers in the world, and Russia received several large orders for calculating and modeling deposits from African countries. Seeing the success of Russian equipment, domestic oil and gas companies began to order settlements from Russian service companies. Supercomputers began to be introduced in production, which led to a decrease in the cost of engines, fabrics, electricity and food by 10-20%.

Realistic forecast

The partial victory of lobbying forces on the part of the government and representatives of science over the domestic bureaucracy allows not only to accept the 25th Rosatom card in the medium term, but also to partially agree to the costs of creating an exaflops supercomputer (30 billion rubles). However, due to the fall in oil prices and a reduction in budget revenues, the program loses the chapter on the creation of a Russian element base (35 billion rubles), as well as some points aimed at training specialists capable of working in supermachines. Instead of the required 30 billion rubles. the industry receives 23 billion, the rest is finished off by co-financing from the commercial sector.

2011: How Russians use supercomputers

In April 2011, the owners of the most powerful supercomputers Russia told CNews^[5]what tasks their computing systems solve. Developers of aircraft engines rejoice that they rarely have to use dead birds for testing, field experiments have also reduced the developers of body armor, and HIV researchers have been able to trace the evolution of the virus under the influence of drugs.

In recent years, Russia has allocated huge amounts of money to create supercomputers. Ordinary people, and sometimes specialists, have a natural question - why are these powerful computing systems used for? CNews talked with the owners of some supercomputers and found out what tasks are being solved on them today.

Lomonosov Moscow State University (MSU)

There are four supercomputers installed at Moscow State University. The most powerful of them - Lomonosov - has a peak performance of 510 Tflops, the rest - 60, 27.85 and 26.76 Tflops. In the Top 50 rating, these systems are located in 1st, 5th, 15th and 26th places, respectively (systems are rated in terms of real performance).

Vladimir Voevodin, Deputy Director of the Lomonosov Moscow State University Research and Computing Center, told CNews that Moscow State University's supercomputer resources are used primarily to support the implementation of fundamental scientific research.

"These are more than 500 scientific groups, of which about 340 are scientific groups of Moscow State University, and the remaining are represented by the institutes of the Russian Academy of Sciences and universities of Russia. At the same time, scientific groups from Moscow State University represent 24 different departments of the university - the faculty or institute, which speaks of an extremely wide range of research carried out using supercomputers, "added Voevodin.

Moscow State University emphasizes "large-scale work on the study of the nature of turbulence, global climate change and the dynamics of the world's oceans, post-genome medical research, design and optimization of complex engineering structures, study of the properties of fluorescent proteins, analysis of the properties of carbon nanostructures, development of cryptography methods, complex polymer studies, subtle methods of analyzing seismic data, mechanisms of galaxy formation and many others."

As examples of specific tasks for which the most powerful supercomputer of Lomonosov University was used, Moscow State University cites seismic data processing, as a result of which the university's scientific teams identified previously unknown natural resource deposits in Sakhalin and Kazakhstan.

At the Lomonosov and Chebyshev supercomputers at Moscow State University, studies are also being carried out on the vulnerability of some cryptographic algorithms in relation to various types of attacks. In particular, work is underway to investigate the so-called hash functions and decompose large numbers into factors.

One of the most famous tasks that was solved with the help of Lomonosov can be called the launch on it of a model for the development of the socio-economic system of Russia for 50 years to come.

Specialists of the IPM named after M.V. Keldysh RAS carried out modeling of aeroacoustics tasks at Lomonosov, for which up to 12,800 cores of the computer system were involved. Such calculations are carried out as part of research aimed at studying the mechanisms of noise generation by aircraft and finding opportunities to reduce its level.

At Lomonosov, heat exchange was also simulated in a mobile phone - the temperature distribution over its surface.

South Ural State University (SUSU)

The peak performance of two supercomputers installed at SUSU is 117.6 and 12.3 Tflops. They are ranked 3rd and 25th in the Top 50.

Dean of the Faculty of Computational Mathematics and Informatics, SUSU, Professor Leonid Sokolinsky told CNews that the distribution of tasks in priority areas of science on their supercomputer resources is as follows:

• 52.2% of tasks are in IT,
• 33.7% - for energy efficiency and energy saving,
• 9.4% - on space technologies,
• 3.5% - on medical technologies,
• 1.2% - on nuclear technologies.

If we take the distribution of tasks by industry, then natural science problems make up 65% of the total flow, engineering - 33%, socio-economic - 2%.

The university uses supercomputers both for its own needs and for settlements on projects of third-party customers. By order of the state corporation Defense industry"," for example, new designs of body armor were being worked out at the university supercomputer, which made it possible to significantly reduce the number of field experiments.

Another task that Oboronprom solved on the SUSU supercomputer was to simulate the mechanics of damage that occur in the human body with local impacts. Previously, for such experiments, either technical plasticine was used, with the help of which it is quite difficult to assess the degree of injury to the real human body, or chest models, which are quite expensive.

The use of a supercomputer made it possible for Oboronprom to significantly reduce the cost of finalizing structures, they say in SUSU. One kilogram of ballistic fabric made of synthetic highly modular material used in body armor costs about $200, and one shot from any weapon at the Russian Center for Testing Personal Protective Equipment at the Research Institute of Steel with speed measurement and registration on technical plasticine - 500 rubles.

By order of one of the knitted factories at the supercomputer in SUSU, modeling of deformation changes in knitted fabrics on the human figure was also carried out. The purpose of this work was to obtain characteristics for the creation of new knitted fabrics that meet the world level in quality. The university's computing power was also used by an investment company to calculate the optimization of the securities portfolio.

Interdepartmental Supercomputing Center of the Russian Academy of Sciences (JSCC RAS)

The JSCC RAS have a supercomputer with a capacity of about 124 Tflops, it takes 4th place in the Top 50.

The power of the supercomputer of the JSCC RAS is provided free of charge to various academic organizations in the order of the general queue, Oleg Aladyshev, chief programmer of the center, told CNews. According to him, the number of users of their system exceeds 1000 people.

In 2010, says Aladyshev, the main areas of research for which the computer system of the JSCC RAS was used were conducted in the field of mathematics, mechanics, physics, computer science and computer engineering, astronomy, chemistry, earth science, biology, biophysics and computer science, all priority areas of modernization of Russia were touched upon.

As follows from the reports of supercomputer users, in the field of medicine, for example, it was used to simulate the microevolution of human immunodificitis viruses. A technology platform has been established to investigate the issue of HIV resistance to antiviral drugs. Biological membranes containing cholesterol and other inclusions were also modeled.

In the field of physics, for example, the supercomputer was JSCC RAS used to study the mechanisms for the transition of slow combustion to detonation during the combustion of pre-mixed gas mixtures in pipes, to study the processes of occurrence and suppression of the knock effect in internal combustion engines. The findings, the report said, provided new material to investigate nonlinear combustion processes and develop new approaches to improving engine efficiency, as well as to develop modern detonation engines.

In the direction of ecology and rational nature management, the supercomputer calculated the propagation of large-scale surface waves in the seas and oceans, and simulated climate and its changes. The system was also used to model global seismicity, develop methods for interpreting electromagnetic monitoring data of the earth's crust in seismically dangerous regions, as well as to model the transfer of radiation in natural environments and solve the problems of global environmental disasters.

Specialists of the IVM RAS and the Institute of Oceanology named after P.P. Shirshova developed and launched on the supercomputer of the JSCC RAS a mathematical model of ocean dynamics, which was used to study the annual variability of water circulation and the level of the Caspian Sea. Using the model, it became possible to prove the existence of subsurface jet currents along the eastern shore of the Middle Caspian and correctly interpret the observation data. Now experts are faced with the task of creating a model of the World Ocean with a spatial resolution that is better than was used in the model of the Caspian Sea.

Institute of Applied Mathematics named after M.V. Keldysh RAS (IPM RAS)

The peak performance of the IPM RAS supercomputer is 107.9 Tflops, it is in 7th place in the Top 50 rating.

Among others, the supercomputer solves problems related to nuclear power - the institute has been cooperating with structures for many years. " Rosatom As the director of the institute, Boris Chetverushkin, told CNews, a significant part of the tasks that load their supercomputer are calculations related to radiation transfer, modeling of atomic reactors.

In addition, according to Chetverushkin, the supercomputer is largely used for aircraft engineering tasks (aerodynamics, simulation of wind tunnels), as well as for modeling oil production, filtration of impurities in hydrocarbons.

Saturn Research and Production Association (NPO Saturn)

Peak performance supercomputer NPO Saturn - 14.3 Tflops, it ranks 28th in the Top 50.

Yuri Zelenkov, IT director of NPO Saturn, told CNews that their supercomputer is mainly used for calculations related to gas turbine engines. As examples of design tasks, he cited the severance of the fan blade, the calculation of combustion processes in the engine combustion chamber, aerodynamic calculations of turbomachines - compressors, turbines, as well as poultry entering the engine.

According to Zelenkov, thanks to the supercomputer, the total design time for products at the enterprise was reduced by an average of 2-3 times and made it possible to abandon experimental design refinement due to its optimization in a virtual environment.

"It is impossible to completely refuse tests, since this is a mandatory part of the certification process in aviation, but all certification tests, including the blade break and the bird throw, now we pass the first time," Zelenkov said.

To calculate the bird getting into the engine, its body is modeled in the form of an elliptical figure with specified properties, but on certification tests, a real seagull corpse is thrown into the engine. Engineers receive dead birds on special farms where birds are raised, Zelenkov explains.

This is how full-scale tests for entering an aircraft engine of a foreign object look like

Nizhny Novgorod State University named after N.I. Lobachevsky (NNSU)

The peak performance of the NNSU supercomputer is 3 Tflops, it ranks 31st in the Top 50. In 2011, a new supercomputer with a capacity of 175.7 Tflops appeared at NNSU, but little is known about its tasks.

Using a system with a capacity of 3 Tflops, the university, for example, conducted modeling of human cardiac activity in order to study the mechanisms for the development of various types of arrhythmias, optimizing the profile of the railway wheel in order to reduce wear on wheels and the railway track.

Rosatom State Corporation

The system installed at Rosatom is the most mysterious in Russia. Its peak performance, as employees of the corporation assure, is 1 Pflops, but it is not listed in the Top 50 ranking. Almost nothing is known about the tasks solved on the petaflopsnik.

In addition to high performance systems, the Rosatom-controlled Federal Nuclear Center in Sarov (RFNC-VNIIEF) also produces personal supercomputers. In 2010, VNIIEF transferred 15 such systems to eleven Russian industrial enterprises.

Together with the hardware, VNIIEF also transferred proprietary application software packages for validation to enterprises. The nuclear center expects that over time their software will allow replacing similar software for numerical modeling from foreign manufacturers.

As an example of the use of their "staff" in Rosatom, they cited joint work with OKB "Sukhoy" to create detailed computer models of large dimensions for calculating an emergency landing with an unreleased landing gear of a new medium-haul passenger aircraft. Superjet-100 In addition, Sukhoi uses mini-supercomputers to simulate the breakage of the fan blade of the new Burlak D30KP gas turbine engine and aerodynamic calculations of the Su-30MKI maneuverable aircraft under given cruise conditions.

At Kamaz, using Rosatom systems, modeling of dynamic deformation of the structure of the KAMAZ-43269 tracking vehicle during explosive loads is being carried out.

AtomEnergoProekt Saint Petersburg conducts design studies of strength properties of the melt localization body during thermal loading. Its results are used to justify the safety of nuclear power plants in the conditions of a hypothetical severe accident accompanied by the melt flowing outside the nuclear reactor vessel.

2010: Production of supercomputers with capacity of 1 Tflops organized in Russia

In total, more than 50 supercomputers are deployed and operate in Russia, which are presented in the national ranking of TOP-50 supercomputers. The leader of the rating is a supercomputer deployed at the Research Computing Center of Moscow State University. M.V. Lomonosov, which in the world ranking of TOP-500 supercomputers takes 17th place. In total, there are 11 Russian supercomputers on the world list.

According to Ministry of Digital Development, Communications and Mass Media the Russian Federation, in 2010, the production of compact supercomputers with a capacity of 1 Tflops was organized in Russia (the first serial batch was delivered to enterprises of high-tech industries - OKBM Afrikantov OJSC, OJSC, OKB Atomenergoproekt Gidropress OJSC, OJSC OKB Sukhoi). Research was carried out on the development of a universal compact supercomputer capacity of 3 Tflops, the mass production of which is scheduled for 2011. In addition, work is being completed to bring the power of the Lomonosov Moscow State University supercomputer to a peak performance of 500 Tflops.

The first versions of domestic packages of imitation modeling programs on supercomputers with mass parallelism have been developed, imitation modeling packages are being adapted at the enterprises of OKBM Africantov OJSC, OJSC Atomenergoproekt"," OKB Gidropress OJSC.

2009: Sales growth of 20%. The main consumer is science

The supercomputer market in Russia shows high growth dynamics - in 2009 the market grew by 20% compared to 2008. At the same time, the main consumers of such services are science (82%), as well as the strategic sector (13%), industry (3%) and the mining sector (2%).

As of July 2009, according to Igor Shchegolev, head of the Ministry of Telecom and Mass Communications, 47 large supercomputer centers were created in science, education and industry organizations. The total performance of these systems, taking into account special-purpose supercomputers, is today 521 teraflops. Shchegolev also recalled that experimental launches of a domestic supercomputer with a capacity of up to 500 teraflops are planned at Moscow State University this fall, and there are also (although still very few) positive examples of the introduction of small supercomputers (up to 1 teraflops) in the commercial sector.

"Despite the fact that we have something to show, we are significantly behind world leaders," Dmitry Medvedev said. - In the list of countries where the most powerful supercomputers are installed, we are in 15th place. Of the 500 supercomputing systems, 476 are known to be occupied by computers made in the United States. "

This situation, according to the head of the Ministry of Telecom and Mass Communications, is explained by insufficient support for the applied use of supercomputers both from the state and from the commercial sector. In addition, the lack of a digital telecommunications resource and often inflated prices for its use affected - and continues to affect - here.

According to Shchegolev, the basic principles of the foundations of state policy in the field of supercomputers and grid technologies should be centralized management and organization of work on the creation of appropriate infrastructure, targeted allocation of funds for the most important projects, state support in priority areas of their use, as well as private-state partnership in programs and projects. The minister stressed that specific projects are already being prepared to increase the capacities of existing centers, as well as to connect all centers into a single network. Other areas in which certain steps are already being taken are the training of specialists and the popularization of such solutions, "perhaps even the inclusion of such solutions in the conditions for providing state assistance for certain large projects."

The topic of stimulating the domestic supercomputing industry was raised by the President of the Russian Federation at a meeting of the Commission for the Modernization and Technological Development of the Russian Economy. Then it became known that 2.5 billion rubles would be allocated to create a supercomputer with a performance of quadrillion operations per second (1 petaflops) at the All-Russian Research Institute of Experimental Physics.

Notes