Soviet Strategic Computing

Today's Russia can hardly be regarded as a world leaders of the computer industry, yet at the beginning of the computer era, the Soviet Union competed on an equal footing with the world's leading computer powers.Since the 1960s, the Soviet computer industry has gradually lagged behind the world level. Differences in the speed of domestic and foreign processors, the quality of peripheral devices, the degree of integration of the element base became more and more noticeable and ultimately led to a loss of Soviet competitiveness.

While computers were in the prototype stage, nuclear power became a reality. By 1948, the USSR had a functioning nuclear reactor; by August 1949, they had exploded a nuclear bomb; and by 1953, they had produced a hydrogen bomb capable of being delivered by an airplane. Nuclear and space research were the Soviet postwar priorities, and the computers that existed to support these highly classified programs were uniquely designed prototypes built for a limited purpose.

What caused the backlog? There were a number of objective and subjective factors, but the most significant of them are two: a low level of technology for manufacturing the main components of computers and orientation to the reproduction of prototypes. It is difficult to say which of these factors was more serious. It must be said that unlike a nuclear or space project, the vital necessity of creating a computer was not realized by the country's leadership.

Microelectronics and computer technology became a key factor in the qualitative development of weaponry in the West, and hence in the achievement of decisive superiority over the Soviet Union. Without question, the USSR's track record with computers compared poorly to that of the West and contributes to Soviet technical and economic woes. It was an ironic state of affairs for the Soviets that their own ideology, which places science and progress on the altar for public worship, had been unable to absorb and exploit the computer.

It was clear that in order to reduce the number of possible variants, it was necessary to apply the mathematical foundations of modeling nuclear explosions, primarily for calculating the power of nuclear charges. Such calculations were organized in Lipan by S.L. Sobolev and at the Department of Applied Mathematics of the Steklov Mathematical Institute (now the Keldysh Institute of the Russian Academy of Sciences) A.A. Samarskii, even before the appearance of the first domestic computers, with the help of computation teams on desktop count- key machines. Even then they proposed effective algorithms for the numerical solution of the equations of mathematical physics, which described the processes of a nuclear explosion. In 1953 the second edition of the monograph by AA Samarskii and AN Tikhonov, "Equations of Mathematical Physics", was published, in which the experience they received was reflected (naturally, without reference to the calculations that served as the source of this experiment).

Thermonuclear calculations were too complex for human efforts alone and so the computer was now a hard necessity for national and international security. Although the Soviets had apparently a developed spy network, their computing lag was a significant variable in holding back their progress.

By 1948, in the initial period of the H-bomb project, experimental data on many determining processes were extremely scarce; no computational capabilities were available. As the authors of the final work on the RDS-6t project noted in 1953, "the joint solution of all equations of this problem, taking into account all the processes occurring in the system simultaneously, is practically not feasible before the development of computer mathematical techniques. Therefore, it was necessary to divide the solutions of three main problems: a) hydrodynamics; b) the kinetics of nuclear reactions and the diffusion of fast particles arising during the reactions; c) radiation."

The theoretical framework for the production of modern computer hardware is found in the science of cybernetics and microelectronics production techniques. During the 1940s and 1950s scientific pursuits in electronics, computers, and cybernetics were officially banned by Stalin - a position dutifully endorsed by the Soviet Academy of Sciences. These subjects, the foundation of today's high technology electronic world, were labeled "bourgeois pseudosciences" and their study was forbidden.

The history of Soviet cybernetics begins in 1953 with the death of Stalin, and when one of the first Soviet computers, the BESM, was completed. The promotion of Krushchev to leadership, even in the early days of the troika, was a boon to Soviet computer science. Krushchev had supported Mikhail Lavrent’ev since 1949 when the scientist showed him the MESM, and thus was not diffident about bringing about computers to the forefront.

Sergei Alekseevich Lebedev designed the Soviet Union's first electronic computer, the Malaya Elektronnaya Schetnaya Mashina or “Small Calculating Machine” (MESM), beginning in 1948 at the Computer Center in Kiev. Only by the end of 1949 the concept of the machine blocks was determined. Then began purely technical difficulties - the very ones with which the Americans had faced several years before. But by the end of 1950 the computer was still built and was in its initial testing stages. After debugging, at the end of 1951, the MEMS was tested and commissioned by the USSR Academy of Sciences Commission headed by Academician Mstislav Keldysh.

Lebedev went to Moscow to head the Institute of Precise Mechanics and Computer Engineering and to design the BESM [Bystrodeeistvuiushcaia Elecktronnaia Schetnaia Mashina, or the high speed electronic calculator], the country's first large computer. The first BESM was completed around 1952. BESM was a binary, 39-bit, 3-address machine with floating point; it has some 4000 tubes and was originally supplied with an acoustic delay-line store. Based on a formulated mix of operations, its speed was usually rated at 7000-8000 opns/sec. In 1953 came the other project, the Strela computer or the “arrow.” Strela was conceived at the Nauchno-Issledovat’elskii Institut Schetnovo MashinostroyeniyaNII Setomash, (SKB-245) and the Zavod SchetnoAnaliticheskhik Maschin or the Calculating Machine Factory. The project was headed by Mikhail Lesechko, Iurii Bazilevsky, Bashir Rameev and Lev Gutenmakher. The massive, elegant but very power-hungry Strela mainframe. Designed and built by the Moscow Plant of Computing-Analytical Machines, Eventually, all seven Strelas were shipped to the Russian Academy of Sciences' Division of Applied Mathematics, where they worked in parallel on nuclear and ballistic missile calculations.

The first programs for the Strela machine, which implement algorithms for the numerical solution of problems of modeling a nuclear explosion, were developed at the Institute of Theoretical Physics of the USSR Academy of Sciences. Although the performance and, most importantly, the reliability of this machine to solve such problems were not sufficient, the first tasks were solved thanks to the virtuosic work of programmers.

David Holloway [Stalin and the Bomb] stated that the “BESM was close to matching the performance of the IBM 701.” Russia sources claim that this could not be further from the truth. The 701 operated 16,000 to 17,000 operations per second. BESM, and for that matter Strela paled in comparison, operating at far lower speeds. BESM was only capable of 600 to 800 operations per second. And the Strela was capable of 2,000 operations per second.

In the 1950s AA Samarskii and AN Tikhonov actively developed the theory of difference schemes, which made it possible to reduce the numerical solution of differential and integral equations of mathematical physics to the solution of algebraic difference equations. These results, published in the Reports of the Academy of Sciences of the USSR in 1956-1959, later used to algorithmize and program the tasks of the Soviet nuclear program on the M-20 and M-220, BESM-6 in the Institute of Applied Mathematics of the USSR Academy of Sciences, the Institute of Atomic Energy, the All-Union Research Institute of Experimental Physics (Arzamas-16), VNII technical physics (Snezhinsk, "Chelyabinsk-70"). The Soviets produced several new computers in 1958, including the M-20 and M-40. The M-20 computer was developed and utilized at the Institute of Precise Mechanics and Computer Technology. It was used in two separate nuclear weapons labs. With 20,000 operations per second, the M-20 was more than able to handle the data for these complex calculations. What was possibly worse for the United States in the arms race was the M-40, which was capable of double the operations per second. Its task was anti-ballistic missile defense.

Work on BESM-2 at the Institute of Precise Mechanics and Computer Engineering began almost immediately after the completion of BESM-I. It proceeded rather slowly, however, and the machine was not completed until early 1959. It differed from the BESM-I in its use of semiconductor diodes and its greater memory capacity. BESM-2 operating speed was 8000-10,000 opns/sec. The BESM machines were intended for scientific and engineering applications, while the M-20 is generally considered to be the "industrial" counterpart of the BESM-2. The M-20 was capable of 20,000 opns/sec (as the "20" indicates). The M-20 has 4500 vacuum tubes and 35,000 diodes.

Lebedev and the Institute of Precise Mechanics and Computer Engineering developed a machine was first called BESM-3, that soon went into production under the name BESM-4. There are reportedly some slight differences between the two, but essentially they are the prototype and production models of the same machine (although several BESM-3 machines were made). The BESM-4 is being widely used and apparently had been manufactured in large numbers. This is be a somewhat simplified account of the development from M-20 tc BESM-4 and M-220. BESM-4 was a transistorized, 3-address, binary machine. Its speed is 20,000 opns/sec.

In 1965, work began on the development of operating systems and automatic programming systems for the BESM-6 computer. In 1967, the USSR appeared supercomputer BESM-6, which could perform 1 million. operations per second. A modernized BESM-6 in 1975 worked as part of the computer complex during the space flight Soyuz-Apollo.

The BESM and M-20 computers were not a family of machines in the Western sense or in the sense of the new Urals. They represent successive design generations, produced by a team headed by Academician S.A.Lebedev, director of the prestigious institute of Precise Mechanics and Computer Engineering of the USSR Academy of Sciences.

By the mid-1960s, Soviet computer designers had only recently begun to develop true computer family groups. There had been in the past several series of very similar machines, in the sense that the same electronic modules were used for each member and input/output functions were based on the same hardware capabilities. But only the Ural line of computers constituted a family of machines in approximately the same sense as the term is used in the West.

In the 1960s, Soviet engineers created several computer lines in Moscow, Kiev, Minsk, Yerevan and Penza. All these machines were incompatible with each other, and each of them had to write all the software anew. Therefore, in 1967 [1969] the Soviet government decided to copy the best samples of Western computers, then to borrow software for them in the West.

But by 1970 none of the Soviet computers in use were third-generation machines; they were first-generation (tubes), and second-generation (transistors) machines equipped with wholly inadequate peripherals and relatively primitive software. The Soviet programmers lagged behind their American predecessors for 6-8 years.

The "make or take" decision haunted the Soviet computer industry. The Soviet government decided to put a stop to these unique Soviet developments and resolved to pirate copies of Western systems instead. As a result, an entire industry’s progress was halted. In addition, the industrial base necessary to make computers has not fully developed. Early Soviet efforts to copy the IBM System 360 were failures.

“This was the worst possible decision,” says Yury Revich, a historian and programmer. “The Soviet government and partly the builders themselves were to blame for the fact that the industry ceased to develop independently. Each group cooked in its own juices and the regime of secrecy made it easier for several technological solutions to be borrowed from Western scientific journals.”

The decision was made at the highest government levels of the Soviet Union to coordinate development and production of a new series of compatible machines with its COMECON partners. This series of machines – the Unified Series or ES – was to be based on the IBM System/360. When the Ryad series of Soviet computers was introduced, each was an exact copy of the IBM-360/370 down to its wiring and English language instruction set. Each Ryad model followed its IBM equivalent model by about 8 years. Soviet models did, however, have serious software problems and were less reliable than the IBM originals. The Soviets illegally acquired IBM 360 and 370 computers from the West in 1972. In 1976, a Riad-40 (IBM 360 clone) was purchased from East Germany by an independent research agency in Washington, DC. Upon close inspection the Soviet Riad computer proved to be far less of a machine than its IBM counterpart of the previous decade. The consequence of the Soviets' policy to replicate the computer systems of the West rather than design their own has been even more damaging in the production of software than of hardware.

The USSR continued to experience serious delays in the development, production, installation, and effective use of its RYAD computers, which formed the cutting edge of the Kremlin's computer modernization program. In the Ninth Five-Year Plan, 1971-75, the USSR and its East European allies produced only 10-15 percent of the anticipated number of RYAD computers, a series of third generation computers modeled on the IBM-360 series. Furthermore, output included only the smaller, less powerful RYAD models, with the final product decidedly inferior to the IBM originals in reliability and compatibility and in the quality of associated input-outputand auxiliary storage devices. Despite this poor track record, the Kremlin pressed ahead in the Tenth Five-Year Plan period, 1976-80, with the development of a RYAD II series-similar to the IBM-370 series.

By 1972 the Soviets had realized the full magnitude of their error and made their first major attempt to close the computer gap. The sleepy Russian village of Kryukovo, within an hour's transit from Moscow, was transformed into the Soviet version of Silicon Valley. Fifty thousand people were assigned to work in the three construction plants, eight institutes, and college which make up the renamed Zelenograd Study Institute.

During the period of detente in the 1970s, the Soviets made deliberate efforts to acquire entire computer production factories from the United States. As relations between the two superpowers cooled, the United States cancelled these programs; nonetheless, the Soviets acquired much of the technology they desired by other means. Consequently, the Edinaya Systyema (Unified System) mainframes, made by the Radio Industry Ministry's facilities in Kiev, Minsk, and Penza, were patterned after the IBM 360- and 370-series computers.

The Soviets did not produce another publicly announced large-scale computer until the late 1970s. Many rumors circulated about several other BESM projects, including one to match the CDC-6600, but no known BESkl-6 successor went into production before Lebedev died in 1974.

Control Data Corporation ventured into a 10 year agreement with Moscow to build a supercomputer; the deal turned into little more than a transfer of US computer technology to the Soviets with a minimum of Russian investment. The American firm was ready to take part in the implementation of the project machine, but the joint work did not work. The White House imposed a restriction on the sale of computer equipment in the USSR, for one, freezing not only trade but also scientific and technical cooperation in this field. As a result, the CDC lost a profitable contract, and in the USSR serial production of the computer PS-2000 was established. The Russian PS 2000 supercomputer, sold to India in 1984, was based on CDC technology. The system was originally conceived as a Soviet-CDC joint project, but the U.S. government forced the American vendor to pull out of the project.

In 1978 the Soviets announced that they had developed two new large-scale, high-speed scientific computer. The announcement, in a Pravda article, indicated that the USSR had made a strong commitment to producing these computers. Advanced computer developments were classified in the USSR, and public announcements had been made in the past only when the Soviets had or were very close to having operational models. The most powerful supercomputers were created at the Institute of Precision Mechanics and Computer Engineering (ITM & VT) , later the Institute of Microprocessor Systems of the Russian Academy of Sciences. There were created powerful electronic machines "Elbrus", later used in space operations, and in the nuclear centers Arzamas-16 and Chelyabinsk-70. The basis of radio electronic reconnaissance equipment "Ural" was two computers "Elbrus" (which the pranksters for some reason nicknamed "El-Burroughs").

Work on the first computer with this name was conducted from 1973 to 1978 in ITMiVT them. Lebedev, supervised these works BS Burtsev, the development was conducted with the participation of Boris Babayan, who was one of the deputies of the chief designer. At that time the main customer of this product, of course, were the military. The machines were developed by the same group that produced the BESM-6, which had been the workhorse of Soviet large-scale scientific computers since the mid-l960s.

The two new machines -- the Elbrus-l and Elbrus-2, previously referred to as VS-l and VS-2 -- were comparable, in computational capabilities, to some of the largest US computer systems available commercially. The Elbrus-l, named after the mountain in the Caucasus, had been fabricated for testing, and a well-developed prototype of the much faster Elbrus-2 also existed. The speed of the Elbrus-l reached 15 million operations per second. The amount of RAM that was common to all 10 processors was, using the notation now adopted, up to 64 MB. Elbrus-2 used a new element base, which allowed to increase the maximum performance to 125 million operations per second. In comparison to a contemporary Western supercomputer, the Cray X-MP, the Elbrus-2’s performance on a mix of applications had been measured to be on par.

It is not known how many El’brus-Is were installed; there were probably at least 10, but possibly as many as 20. Only three or four El’brus-2s are in operation. Unlike such general-purpose computers as the Soviets Ryad models, large-scale scientific computers were not intended to be produced in large numbers. Because of their size, design, high performance, and cost, their application is limited to complex military and technological problems in such areas as air defense, nuclear weapons design, aircraft design, seismic analysis, and long-range weather forecasting. "Elbrus-2" computers were operated in nuclear research centers in Chelyabinsk-70 and Arzamas-16; finally, this complex, since 1991, was used in the A-135 ABM system, as well as at other military facilities of the country.

In a remarkable private conversation with a US journalist and former arms control official, Marshal Nikolai Ogarkov, First Deputy Defense Minister and Chief of the General Staff, interpreted the real meaning of SDI: "We cannot equal the quality of U.S. arms for a generation or two. Modern military power is based on technology, and technology is based on computers. In the US, small children play with computers.... Here, we don't even have computers in every office of the Defense Ministry. And for reasons you know well, we cannot make computers widely available in our society. We will never be able to catch up with you in modern arms until we have an economic revolution. And the question is whether we can have an economic revolution without a political revolution." [Leslie H. Gelb, "Foreign Affairs: Who Won the Cold War?," New York Times, 20 August 1992, p. 27. Gelb held this conversation with Ogarkov just days after Reagan's SDI announcement, but he did not report it until 1992.]

A 1984 assessment of Soviet computer production by Director Yury Nesterikhin of the Soviet Institute of Automation and Electrometrics in Novosibirsk noted quality control problems and said that the domestic prices for Soviet microprocessors were 10 times the price of microprocessors produced elsewhere in the world. Yevgeniy Velikhov, Vice President of the Soviet Academy of Sciences, reminded planners of what happened with Soviet-made electronic calculators: "The calculators have been gathering dust on store shelves for much the same reason as digital watches do; they are subject to breakdowns, no one knows how to fix them, and their batteries are rarely available."

By the mid-1980s Soviet computers typically were not very powerful. The workhorse for Soviet computation, the BESM-6 - a second-generation (transistors-on-boards) scientific mainframe first shipped in 1965 - was comparable to widely used US personal computers) and had severe reliability problems. Existing Soviet computers for economic data processing were crude by US standards. Most Soviet machines were designed for the military. In education computers can only be realistically used under repair free operating conditions. If a machine breaks down it cannot be repaired on the spot but must be replaced by a new one.

A 1985 report announced that USSR computer producer Elektronorgtekhnika in Kiev had begun producing an equivalent of the VAX- I I computer. This was Digital Equipment Corporation's follow-on technology to the PDP-11 of the 1970s. At the supercomputer end of the spectrum, the Soviet BESM-6, a collection of imported computer parts, was thought to perform 10 million instructions per second (mips). Western supercomputers, of which 160 mostly US-built machines exist (the other primary producer being Japan), were capable of between 700 and 1,000 mips. In practical terms, this meant that problems requiring 48 hours to compute on the BESM-6 will be finished in approximately six minutes on the Cray-lA, US supercomputer.

Although this class of computer has been found to be invaluable in the West for many important high-fidelity modeling and simulation tasks, and stringent and generally effective Western export controls created strong incentives for indigenous Soviet production, the Soviet Union had not developed the ability to build machines of this class. When the Soviet Academy of Sciences established a Department of Informatics, Computer Technology, and Automation in 1983, one of its top priorities was supercomputer development. Machines with appropriate levels of performance were designed and built in prototype, but could not be built in quantity because the industrial ministries responsible for production could not meet the Academy designers' specifications.

By the middle of the 1980s, the gap between Soviet systems and Western models in performance, quality, and numbers produced had become extreme, particularly in the field of supercomputing. The CRAY-1, with a theoretical peak performance of 160 megaflops, had entered production a decade earlier. The most powerful Soviet machine in volume production, the ten-processor El'brus-2 with a theoretical peak performance of 94 megaflops on 64-bit operands, entered series production in 1985.

In addition, the general-purpose computer "Elbrus 1-KB" was also produced; the creation of this computer was completed in 1988. Until 1992, 60 such computers were produced. They were based on the Elbrus-2 technologies and were used to replace obsolete BESM-6 machines.

Elbrus-3 was never launched into mass production. The only working copy was built in 1994, but at that time it was not needed by anyone. Logical continuation of work on this computer was the emergence of the processor "Elbrus 2000", also known as E2K. The Russian company had big plans for the serial production of this processor, which was supposed to go to the series simultaneously or even earlier than Itanium. But due to the lack of the necessary amount of investment, all these plans were not implemented and remained on paper.

In 1996, Russia expressed a strong desire to obtain high performance computers from the United States for use at its nuclear weapons laboratories. The Convex SPP 2000 computer was more capable than any computer known to have been in use in Russia at that time. A According to the Russian Minister of Atomic Energy, such computers were needed to help Russia maintain its nuclear stockpile, particularly in light of the CTBT prohibiting future nuclear explosions. Russia attempted to obtain high performance computers for its weapons laboratories for “civilian purposes” from two U.S. manufacturers. The manufacturers, in compliance with the export control laws and regulations, sought an export license for the transaction, but the applications were eventually returned by the Commerce Department without action.

The Clinton Administration determined that it was in the U.S. interest to cooperate with Russia on the safety and security of their nuclear weapons stockpiles, but within certain specific boundaries. Pursuant to this policy, discussions were held with the Russian Ministry of Atomic Energy (MINATOM) and other officials on the possibility of undertaking cooperative projects under a CTBT. Department of Energy officials said that the policy boundaries for potential cooperative projects are that they would be unclassified, and most importantly, they would not enhance the performance of Russian nuclear weapons or contribute to Russian nuclear weapons design.

Subsequent to the executive branch’s decision to return the license applications without action, press reports began to circulate here and in Russia that Russia had obtained several high performance computers from U.S. companies, apparently without an export license. Press reports indicated that MINATOM told one of the companies that sold them a computer without a license that the computer would be used for modeling of earth water pollution caused by radioactive substances.

However, MINATOM officials stated that the computers will be used to maintain the Russian nuclear weapons stockpiles and the Minister of Atomic Energy indicated that the computer would be used to confirm the reliability of Russia’s nuclear arsenal and ensure its proper working order under the terms of the CTBT. Because the computers Russia obtained use a technology known as parallel processing, a number of processors can be added to increase their performance.

From 2007-2010, when the government started to actively finance the sciences after a 15-year hiatus, Russian and Belarusian scientists jointly created the SKIF supercomputer series (SKIF is the Russian acronym for SuperComputer Initiative Phoenix). Another supercomputer, the AL-100, was launched in 2008. Its peak productivity reaches 14.3 Tflops. The AL-100 comprises 336 Intel Xeon 5355 processors and has 1,344 GB of RAM. The Lomonosov supercomputer was created in 2009.