Sergey Oboguev (oboguev) wrote,
Sergey Oboguev

Посетивший в конце 70-х СССР классик программирования Э. Дейкстра сказал в своём публичном выступлении в Большом зале Академии наук в Ленинграде, что он считает крупнейшей победой США в холодной войне тот факт, что в Советском Союзе производятся компьютеры фирмы IBM.

. . .

Финалом жизненного цикла ЕСовских мейнфреймов стала их массовая скупка в целях извлечения из них драгоценных металлов. В составе комплектующих машины имелось несколько десятков грамм золота и порядка одного или нескольких килограммов серебра. В период с 1990 по 1993 годы свыше 90% эксплуатировавшихся в СССР мейнфреймов ЕС ЭВМ оказались демонтированы и утилизированы в этих целях.

из статьи S.V. Klimenko, Computer Science in Russia: A Personal View:

In 1959, John von Neumann visited the Soviet Union. On his return to the United States, he declared that Soviet computer science was at least one year behind the Western one. Apparently it was an underestimate, because one may state that the lag of the Soviet Union was more like a few years. According to estimates, by the late 1960s, the lag of the Soviet computer industry was already five years, though it was not considered to be critical. The crisis broke out in the 1970s when IPMCT (ИТМВТ) failed to turn in for commissioning the new Elbrus computer (that was to replace BESM-6) by the appointed time. The commissioning was first planned for 1972, but it was not put into serial production until the early 1980s.

But still, the true catastrophe was caused by the decision of the country’s leadership to reorient the computer industry for the production of obsolete IBM-like series computers. In the late 1960s, the state program to advance the computer industry for the 1970s was under development. There were two competing projects:

1) cooperation with the British company ICL that offered to set up joint production in the Soviet Union and was ready to transfer technology and know-how versus

2) orientation to manufacturing of computers completely compatible with the IBM 360.

The most eminent hardware and software designers advocated the first trend. But strange as it might seem, the industry and the Research Institute for Computer Technology (RICT - НИЦЭВТ) supported the second one. Of course, it was the industry that gained victory. As a result, the Soviet computer industry immediately was left behind for 10 more years. Ten years was exactly the time it took RICT and the Minsk plant to put into serial production a new model of ES-1020 (with a memory of 512 Kbytes and a speed of 100,000 ips). Hence, by the early 1980s, the Soviet computer industry was already 20–25 years behind the Western one. Not only now (20 years later) but even at that time, the decision seemed disputable. But still, the decision was made at the top level as a long-term state program.

. . . . .

BESM-6 and Elbrus 1 KB

In 1973, the preliminary and conceptual designs of a high-productivity computing system, BESM-10, whose architecture was similar to that of AS-6, were done. The project envisioned raising the power of the central processor to 10–15 million instructions per second (MIPS), raising the exchange rate and increasing the throughput of the input/output subsystem of the peripheral machine PM-10 and raising the throughput of all adapters and multiplexers. But in the same year, Burtsev replaced Lebedev as the director of IPMCT. As a result, the BESM-10 project was shut down, and the development of the Burroughs-like computer Elbrus 1 began. The throughput of the CPU of this computer was at a level of 1.5 MIPS, i.e., the same as the AS-6, but it should be noted that it was attained only in 1979. Elbrus 1 could incorporate from one to 10 CPUs, i.e., the total throughput could be from 1.5 to 12 MIPS. The maximum size of the memory was up to 1,024K 64-bit words. The peripherals were from the ES computer (IBM 360). To ensure smooth transfer of the software, it was envisioned to use the integral version of BESM-6, the so-called SVS-1, as one of the CPUs of the Elbrus system. The basic programming system is the autocode El-76, a high-level language resembling Algol-68, with many language structures maintained at the hardware level. A few Elbrus computers were manufactured—this means that there was no real progress in the project and that the progress was evidently inconsistent.

Elbrus 1 DB

The pressure from influential users made IPMCT revitalize the work on developing the integral version of BESM-6, originally named SVS-1, and acquiring the trade name Elbrus 1 DB. M.V. Tyapkin supervised the work. (He had participated in the development of BESM-6.) Elbrus 1 DB was marketed as an independent computer, it had 80 Mbytes of 64-bit memory, its speed was about 6 MIPS and was furnished with 200-Mbyte exchangeable disks and ES-compatible magnetic tapes. The computer occupied two cabinets and had a water-cooling system. In one of its operating modes, the computer is completely compatible with the BESM-6. The history of developing the operating system for Elbrus 1 DB, named Feliks, is interesting. It was the first operating system developed exclusively with a high-level language, Yarmo. The work was carried out at the Novosibirsk branch of IPMCT under the leadership of the brilliant engineer and talented programmer F. Tsang. It was supported by academicians Yershov and G. Marchuk, then the president of the Soviet Academy of Sciences. The work on the operating system had been finished a few years earlier than the first computer was manufactured. The author of the most popular operating system, Dispak, for BESM-6 developed its version for Elbrus 1 DB and for Unix. Dispak was developed at the Novosibirsk branch of IPMCT. It is interesting to note that a prototype of a desktop (personal) version of BESM-6 had been manufactured at JINR (Dubna) a few years before Elbrus 1 DB appeared.

Elbrus 2

After the failure of the Elbrus 1 project, the activities in this area were still going on at IPMCT, with Elbrus 1 interpreted as an experimental computer for refining the ideas, methods, and technology of production and for developing the software of the “real” Elbrus 2. The major difference between Elbrus 1 and Elbrus 2 was the faster processor of the latter, having a total throughput in the 10-processor configuration of 125 MIPS. Elbrus 2 was put in serial production—around 10 computers were manufactured. It was an extremely expensive computer. The components included very delicate elements, e.g., coaxial cable used for atomic submarines. (I know an example of when the Elbrus 2 was supplied to a customer and then was brought back because that cable was faulty.)
  • Post a new comment


    Anonymous comments are disabled in this journal

    default userpic

    Your reply will be screened