The birth of the Soviet missile defense system. The end of the Soviet computer program
During the Korean War, Watson entered into contracts with 18 agencies for the supply of computers, which received the patriotic name of the Defense Calculator Model 701. Since 1955, shipments of the Model 702 began, then its improved version of the Model 705, later the entire line of 700s was equipped with memory based on ferrite rings ...
The excellent architecture of the 700 series is also evidenced by the fact that it survived the hardware change from tubes to transistors, evolving into the 7000 and being released for several more years. Since 1955, the number of installed 700s surpassed the number of machines installed by Remington Rand for the first time.
The older factory in Endicott began producing the equally successful Model 1954 for small and medium-sized businesses in 650. More than a thousand of these computers were sold, thus the Model 650 can be called the first computer to be mass-produced (the release of only one 650th model exceeded the number of all computers in the USSR by an order of magnitude in 15 years, just be aware of this).
Due to competent marketing, large financial investments, connections in the US Government and the experience of mass production of complex machines in the mid-50s. IBM has achieved a dominant position in both segments of the computer market.
One of the indicators of success was the fact that in 1956 their technique was already used to predict the results of the elections. The scientific computer market had a lower potential capacity than the business computer market, so even the successful sales of UNIVAC and its well-known brand name failed to improve the situation at Remington Rand.
And in 1956, it was bought by the already familiar Sperry Gyroscope, forming Sperry Rand, and the ERA and Eckert-Mauchly divisions were merged into Sperry UNIVAC.
The 700/7000 series by this time consisted of 6 lines, and they were practically incompatible with each other, neither software nor hardware (I mean the processor and RAM, the peripherals were compatible), and in addition was divided into two generations - the 700 tube and the transistor 7000th.
As we can see, in those wild and crazy years, no one yet mastered the art of methodically competent machine design, not even IBM. At the end of the 1950s, they had:
Above - IBM 704, IBM 709, IBM 7094, below - IBM 702, IBM 7080 (photo https://en.wikipedia.org, http://www.columbia.edu, https://www.ibm.com)
First (36/18-bit words): 701 (Defense Calculator).
Scientific (36-bit words): 704, 709, 7090, 7094, 7040, 7044.
Senior commercial (variable length words, string type): 702, 705, 7080.
Junior commercial (variable length words, string type): 1240, 1401, 1420, 1440, 1450, 1460, 7010.
Decimal (10 bit signed BCD): 7070, 7072, 7074.
The only supercomputer (64-bit words): 7030 Stretch.
In addition, the 700 line was not included, but appeared simultaneously with it, and was peripherally compatible with the IBM 650 Magnetic Drum Data Processing Machine and the unique IBM 1958 RAMAC (Random Access Method of Accounting and Control) released in 305, the first computer with a hard disk.
NORC
Also, even before the initiation of the Stretch project, IBM is building the Naval Ordnance Research Calculator (NORC) supercomputer for the US Navy Bureau of Ordnance.
The NORC was a very curious machine and didn’t fit into the IBM hardware of this era at all. It was a bizarre blend of scientific computer concepts as understood by academics in the early 1950s, ennobled by IBM technology.
In that part of architecture that had roots at Columbia University (and the restless Wallace Eckert acted as the chief architect of NORC, this was his last work for IBM), he was the ideological successor of SSEC and the closest relative of BESM, something like Lebedev's machines would have been built if he supported by a powerful corporation (this proves once again that from academic scientists, system architects are like a ballerina from a hippopotamus, well, it's not a royal job to think about users). However, no one reasoned Lebedev, and Eckert's ideas were significantly cultivated by a group of experienced engineers of the corporation, as a result, the hybrid of a hedgehog with a snake turned out to be much more elegant than in the USSR.
NORC was the first supercomputer in the sense that for the first time in the world it was originally built with the goal of surpassing all other machines in power and for the first time in the world enough computers were already made to compete with.
Accelerating to 15 KIPS, he fulfilled his task (which backfired even in the USSR, as we remember, the vindictive Bruevich wrote in a review of BESM that it was no good for NORC, and Lebedev was given a prize that time). Its architecture, however, was so strange that none of the NORC concepts were subsequently applied directly to IBM machines.
What was special about him?
Decimal arithmetic, both real and integer (BCD-code, 16 decimal digits, 64 bits + 2 bits of error correction modulo 4). The word could store a 13-digit signed number with a 2-digit index, or one instruction. Three-address (hello Lebedev!) Instructions, 64 in total, two general and three index registers - a scheme a bit similar to the CDC 1604. The whole machine was assembled from replaceable modules in the spirit of IBM, only 1 blocks, 982 types, but half of the schemes used only six of them.
In general, NORC was a BESM of a healthy person and at the same time made it clear what Lebedev would have built if he worked at Columbia University (as well as the fact that he would not have been given anything else to build).
At NORC, Eckert's career as a systems architect came to an end, IBM was happy to use his services as a mathematician and physicist, but he was no longer allowed to play with developing machines, since his knowledge of computer architecture was stuck in the early 1950s.
Nonetheless, NORC did have an important impact on the industry.
During its development and assembly, IBM engineers were trained in practical concepts of working with electrostatic RAM, timings and more, which were then used in the 701 series.
NORC also became the first machine in the world to include a channel coprocessor, which in many ways made it possible to squeeze this performance out of lamps.
The architecture of magnetic drives also migrated to 701. This idea was considered very successful and migrated to Stretch, and then to S / 360. At the presentation, NORC showed its power, within 13 minutes calculating the number π to the 3-thousandth decimal place, which in those years was a world record (the idea was proposed by von Neumann, he really wanted to make sure that all the numbers are random).
Article about NORC in the New York Herald Tribune, 1954. NORC Presentation - All the country's top officials associated with computing: Wallace Eckert, father of American nuclear weapons Oppenheimer, Columbia University Vice President George Pegram, von Neumann himself, and Thomas Watson-Sr. NORC itself in all its glory (http://www.columbia.edu)
So what are the achievements of the 700/7000 series that have eclipsed the glory of UNIVAC and pushed all other companies into the shadow of IBM?
In 1954, the Joint Chiefs of Staff requested a machine comparison - the world's first test of different architectures. He showed that the IBM 701 is slightly faster, but the ERA 1103 performed I / O operations much more efficiently, due to the I / O coprocessor, this idea will firmly sink into IBM's soul and will be embodied in Stretch. In addition, this test has drawn attention to commercial computers and influenced the openness and deregulation of a formerly secret industry.
The 704 series became improved and incompatible in terms of commands. As we already said, it was supplied in large quantities to universities, the classic FORTRAN and LISP languages were first developed specifically for it. In addition, the Smithsonian Astrophysical Observatory calculated the orbit of the first Soviet satellite on it. From an architectural point of view, this machine was notable for being the first mass-produced computer with hardware support for floating point calculations and index registers, which greatly accelerated the work and simplified programming.
We have already talked about Stretch, as well as the fact that 7090 and 7094 were purchased by NASA.
IBM 1401
Finally, do not forget about the junior commercial line, the transistor IBM 1400. The 650 and 704 models brought the company fame, and with the IBM 1401 Data Processing System, the tabulators began to decline.
The combination of functionality and relatively low cost of the 1401 has allowed many firms to embrace computer technology, and its popularity has helped IBM become a market leader. Remington Rand was unable to offer anything similar.
IBM was the first to realize that profit is made not by piece super-expensive installations, but by a simple mass product. For the first time, the combination of cost, reliability and functionality made computers very attractive to many consumers.
In a way, the 1401 was too good, as Watson Sr feared, consumers one after another began returning their rented tabs to IBM to take on a new miracle. This caused a lot of short-term problems for the company, but it decided to be patient and was not mistaken.
Magnetic RAM, transistors, advanced software, and printers were huge breakthroughs for the 1400 series, and each gave a big advantage in the market, and when combined with the low price tag, it was a killer combo.
Sales of the 1400s outnumbered tabs by a factor of ten and brought in super profits.
The Model 1401 became the most successful computer of the 60s, with more than 12 mainframes sold, although its incompatibility with the 7xx line became a major problem. This caused a lot of inconvenience, both for customers and for IBM itself.
The company had to train service personnel and provide software support for each individual system (again, in the USSR, developers in most cases, to put it mildly, did not spit on users). This led to the creation of a special group SPREAD (System Programming, Research, Engineering and Development) to investigate the possibility of creating a new universal and compatible line of computers.
The 70xx and 14xx series of computers made IBM widely known, and sales in a little over six years doubled from $ 1,17 billion in 1958 to $ 2,31 billion in 1964, growing at a rate of 30% annually.
According to Datamation magazine, in 1961, already 81,2% of the computer market belonged to IBM.
IBM's comprehensive approach also included software. For the first time, completely free of charge, IBM included software packages that met most of the customer's needs, rather than leaving software development to users. This was critical because the software packages saved a significant amount of time and money in in-house development and allowed organizations without programmers to finally benefit from computers.
Volunteers help rebuild the 1401 mainframe for the Computer Museum in Mainin View, California. It is a pity that the most interesting samples of Soviet cars (from "Kronos" and BESM-6 to boards "Elbrus-2") can be found much more often and in better safety in Western museums, and not in Russia (https: //en.wikipedia .org, https://www.computerhistory.org).
SHARE
Customers and users of the IBM 701 formed the first in 1955 in Los Angeles stories computer technology group of users called SHARE, which was also the first organization in the computer industry to pursue standardization. That's when IBM opens its first program pre-testing center.
Later, it became the largest forum for the exchange of technical information on programming languages, operating systems, database systems and user experience for corporate users of small, medium and large IBM computers such as the S / 360, S / 370, zSeries, PSeries and xSeries.
Initially, IBM distributed its operating systems in source code, and system programmers usually made small local additions or modifications and exchanged them with other users.
The SHARE library, and the distribution process it supported, was one of the main sources of open source software.
In 1959, the group released the SHARE Operating System (SOS), originally for the IBM 709 computer, later ported to the IBM 7090. SOS was one of the first examples of "co-production" now widely used in the development of open source software such as Linux.
In 1963, SHARE collaborated with IBM in the development of the PL / I programming language as part of the 3x3 group. The organization still exists today, publishes a newsletter and holds two major educational meetings annually.
In 2005, there were over 20 members of this group, representing some 000 IBM corporate customers.
In the USSR, there was nothing close to such a model of working with software.
"IBM and the 7 Dwarfs"
IBM's success has been fueled primarily by serious R&D, which has resulted in the firm becoming the owner of key patents.
Their costs were increased from 15% of income in 1940 to 35% in 1950 and 50% in 1960. Since 1960, IBM's science budget has surpassed the US federal science budget!
Second, as Watson bequeathed, customer and sales focus.
The company had a wealth of experience in the sale and service of complex systems that competitors did not have. In addition, IBM did not ignore any market or group of potential customers, as did many firms that focused exclusively on powerful scientific or military computers.
As a result, by the end of the 1950s, a situation arose on the computer market that was later called "IBM and the 7 Dwarfs."
In addition, even before the creation of the S / 360, several important events happened in the life of IBM.
They got involved in two of the largest infrastructure projects of the 360th century - the creation of the SAGE system and the development of NASA's MCC at Cape Canaveral for the Apollo program. Both projects were huge successes and brought the company a huge amount of money, respect from the government and invaluable technical experience, much of which was then applied to create and promote the S / XNUMX line and all subsequent products.
We wrote about the NASA project in the previous part, SAGE should be written separately, because the topic is completely immense.
We only note that it also grew out of the SABER (Semi-Automatic Business Research Environment) system for American Airlines, created by IBM in 1962, originally operating on 7090 mainframes. lines, she laid the foundations for all this kind of technology. If not for SAGE and SABER, readers would not order pizza through mobile applications now.
Another great breakthrough was the creation of FORTRAN in 1957.
IBM revolutionized programming by creating a scientific formula translator so convenient that it became the standard for several generations of scientists, and libraries in this language are still used in some places.
Step 4. Triumph of the Unified System
The first mainframe of the most famous line of IBM came out in 1964, and the revolution he started was comparable to Hollerith's tabulator.
Like the Intel 8086 processor, this machine spawned a long line of descendants and became the standard for many years to come. The only difference is that Intel initially did not predict great success for these particular processors and developed them, in fact, as a temporary measure, which by chance became famous. Hence, at least two attempts by the company itself (iAPX 432 and Itanium) to bury the not very successful x86 architecture, which, however, ended in even greater failure.
IBM originally wanted to develop a machine for decades, and they succeeded. For its presentation on April 7, 1964, IBM held 77 press conferences in 15 countries around the world, making, according to the head of the firm Thomas Watson Jr., "the most important announcement in the history of the company."
What was his claim based on?
Professional approach to computer architecture development - IBM took into account the failure of Stretch and in advance specified in detail everything related to hardware and the system of commands in a set of developer guides, the most important of which were "IBM System / 360 Principles of Operation" and "IBM System / 360 I / O Interface Channel to Control Unit Original Equipment Manufacturer's Information manuals ". It was with the S / 360 that such specifications became standard.
The first industry standard architecture: a compatible hardware and software line of 6 machines of different performance and price, and 40 peripherals for every taste and wallet with the possibility of upgrading.
The architecture of the machine is specially designed to be universal - both traditional features of scientific machines are supported (full-fledged real arithmetic FORTRAN), and new ones for business are introduced (decimal arithmetic, COBOL).
IBM's early computers originally used transistors licensed from Texas Instruments. Subsequently, they decided to produce all the electronic components themselves, so as not to depend on external suppliers and to ensure the lowest possible prices. For the S / 360, a universal standard for GIS and SLT (Solid Logic Tecnology) boards was developed.
To reduce the cost of production of the most expensive component - RAM, for the first time in world practice, a plant was opened in Japan. The factories were subsequently relocated to Hong Kong, further driving down costs. IBM's competitors have followed suit and have also begun to gradually relocate manufacturing facilities to Asia.
For the first time, widespread use of hardware virtualization: a technology that came to desktop processors only in the mid-2000s, since the S / 360 era, it has become a trademark of mainframes and the main reason for their incredible flexibility and reliability.
Replaceable processor firmware made it possible to effectively emulate legacy IBM computers - and thus another basic mainframe rule was born, full compatibility. Until now, COBOL programs written under S / 360 can run on z / 10, released in 2008.
An incredible number of technological innovations: microcode in the processor, 32-bit general-purpose registers (instead of the ancient "register-accumulator" scheme, and this archaic scheme was used at that time even in CDC supercomputers!), A huge amount of RAM at that time (16 MB , PCs were able to address such a volume only in the late 1980s, older S / 360 models could address 4 gigabytes, not every home in 2005 had that much RAM!), I / O coprocessors, dynamic address translation (DAT), time sharing, 64-bit real registers, write protection, multiprocessing support, etc.
Amazingly, the S / 360 was the first to match the word length, adder and address (although different combinations of their lengths could be used).
Unfortunately, to assess the incredible progressiveness of this solution, you need to be an assembler programmer, but keep in mind that the legendary BESM-6, for example, had an adder capacity at least a multiple of the command length (48 and 24 bits), but the address was not something that was not multiple, but not a power of two (15 bits) at all, and a byte was six bits! Machine programming was hell for her.
IBM spawns new standards: nine-track tape, EBCDIC code table; 8-bit bytes (now this may seem surprising, but during the development of System / 360, for financial reasons, they wanted to limit bytes to 4 or 6 bits, the option of bytes with variable length and bit addressing, as in the IBM 7030, was also considered) and byte memory addressing; 32-bit words (and in general, the standard 8, 32, 64 bits); IBM architecture for real numbers (actually a standard for 20 years, before the introduction of IEEE 754) and hexadecimal constants. The hexadecimal number system, widely used in the S / 360 documentation, supplanted the previously dominant octal one.
All this made the line extremely resilient (its descendants are still being produced today), amazingly successful commercially (in the first month alone, IBM drowned in more than 1100 orders, many companies bought seats in the queue for the delivery of new computers) and incredibly flexible (these machines worked everywhere - from the Apollo software to the accounting departments of IBM itself).
In one fell swoop, the company swept away all competitors from the market.
A few years later, RCA and GE dropped out of the mainframe manufacturers, then Honeywell first merged with Bull, and then went bankrupt, CDC could not stand the competition by the end of the 1980s, and only UNIVAC and Burroughs, united in UNISYS, were able to resist the IBM empire. If the S / 360 failed, IBM would disappear with it - they invested wild money in the construction of six factories around the world, hired an additional 50 thousand employees, the program combined about 2 other projects.
Initially, we were faced with two risky moments at once. Even one of them was enough to keep you awake at night ... First, the task was to coordinate the design of hardware and software for the new line. We had teams of engineers all over America and Europe working on six new processors and dozens of new peripherals at the same time ... but in the end, all this hardware would have to be put together. Secondly, the software. To support System / 360, hundreds of programmers have written millions of lines of code. An unprecedentedly challenging project. Never before have our engineers been under such strong pressure,
- recalled Thomas Watson Jr.
Here is the percentage of installed base of electronic data processing equipment of major suppliers in the United States (1955-1967):
As a result, of the 10 billion total cost of installed computers in 1964, the "gnomes" produced 30%, and IBM - the remaining 70%.
Finally, we can note the latest and greatest innovation of the company, introduced everywhere in the West precisely after the release of S / 360 - a scientific approach to managing the development of not only hardware, but also the software part, what is now called software engineering.
A revolutionary machine demanded a revolutionary operating system, and OS / 360 was supposed to be just that: multiprogramming, virtual memory and virtual machines, working with multiprocessor configurations - this is not a complete list of innovations inherent in its architecture. The operating system had to work on all models of the line, so the configurations varied from 16 KB of RAM to 1 MB, and the speed of operation - from several thousand operations per second to half a million.
Also, the operating system had to satisfy the needs of all programs, starting with complex mathematical calculations that almost did not use external drives, and ending with simple analogues of a DBMS, which were completely based on I / O operations.
But if IBM had already realized that it was impossible to design hardware anyhow, then no one had ever written such complex programs until then, and there was no understanding that the principle of competent design should be observed for software as well.
As a result, a huge development team frantically tried to write, dock and debug millions of lines of code in pure assembler, several months after the machine itself was completely ready, waiting only for the OS for it. The hardware part was ready to go on sale, and a stable and reliable version of OS / 360 was never born, in addition, the final cadaver did not want to fit into the memory of younger models.
To save the day, OS project manager Frederick Phillips Brooks, Jr. ordered the release to be split into three parts with the promise of further upgrades. This is how BOS / 360 (Basic OS), TOS / 360 (Tape OS) and the famous DOS / 360 (Disk OS) appeared - the most powerful version of the OS did not fit entirely into RAM and was not suitable for booting from a slow tape, so we had to use a hard disk. OS / 360 itself required millions of man-hours of work, but its complete and complete version never saw the light of day.
IBM's enlightenment was as complete as in the story with Stretch - Brooks realized the omissions of his team and in 1975 released the developer's bible, The Mythical Man-Month: Essays on Software Engineering (immediately translated into Russian, however, in the conditions of the USSR, it was useless).
Thus was born the second classic discipline of Computer Science - software development.
In addition, the S / 360 architecture formed the basis of the most famous American avionics series, the IBM System / 4Pi. The name is also given here with a hint - in the sphere of 4π steradians, as well as in a circle of 360 degrees. This platform was also designed to fill all possible niches of onboard computers, and it did it, S / 4Pi machines were used in F-15 Eagle, E-3 Sentry AWACS fighters, Harpoon rockets, NASA Skylab ships, MOL and Space Shuttle and a huge number of other aircraft.
It consisted of 4 basic models: TC (Tactical Computer), the size of a briefcase for controlling missiles, helicopters and satellites, weighing about 8 kg; CP (Customized Processor), medium power for aircraft, radar and mobile battlefield systems weighing 36 kg and 21 kg in the CP-2 version; EP (Extended Performance), for applications requiring large amounts of data in real time, such as manned spacecraft with a crew, warning and airspace monitoring systems and command and control systems, weight 34 kg.
All models used a command architecture that was a subset of the S / 360 (for example, EP - S / 360 model 44), and applications for them could be developed directly on the IBM mainframe. The Skylab station used the TC-1 model, 16-bit words and 256 KB of RAM. The flagship model AP-101 was 32-bit, used firmware like a large mainframe, and could address 1MB of memory.
This model was used in the Space Shuttle (AP-101S), aircraft B-52 and B-1B (onboard local area network of 8 computers!) And many others. The AP-1, a bit simpler, was in the F-15. An old Gemini computer produced 0,007 MIPS, while the AP-101S could accelerate to 0,48 MIPS, half the power of a BESM-6 in a small suitcase!
Shuttle used an architecture in the form of a network of 5 AP-101s, each with its own channel coprocessor for 24 buses, an idea borrowed from mainframes. Four computers worked in parallel to achieve fault tolerance, the fifth was a backup, and its software was not a copy of the rest, but was developed and tested separately to ensure greater reliability.
Navigation and control software was written in a special NASA language - HAL / S, and the OS was in assembler. The aircraft software was written in JOVIAL.
In the USSR, such an advanced concept was simply impossible - in our country, all military and space computers were developed entirely on a specialized basis, all were unique and incompatible with each other. CADC flew off until 1980, and was forgotten, IBM System / 4Pi, on the other hand, managed to visit space and worked until the mid-1990s, of course, gradually being modernized.
So here is the journey that IBM took in 1880-1965, 85 years of hard work, technical innovation, business and education development and the largest infrastructure projects of the era - SAGE, SABER and Apollo - culminated in the creation of the absolute architectural masterpiece, System / 360.
It is striking that of the 5 most important conceptual innovations in the entire history of computers - mainframes, personal computers, wearable electronics, graphics processors and neurocomputering, IBM is responsible for three and a half (in 3 cases out of 5 they presented a reference product for the industry, in the case of neural networks - studied a lot of the theory of the issue, and the first experiments with AI were carried out back in the 1950s on machines of the 700 series).
We have already seen how the path that the USSR went through radically, literally every day, differed from the path of the IBM company.
Hence the simple answer to the question - could the Union in 1965 present an absolutely alternative architecture that would be so successful?
The simple answer is no.
To beat IBM, it was necessary to start in the middle of the XNUMXth century, when the USSR was not even in the project, and to build the whole history over the years in a completely different way.
By 1965, it was realized that at such a pace as it is, we will be catching up with IBM for another 50 years, and thousands of computers are needed right now.
For 15 years of informatization, no more than 1 computers of approximately 500 absolutely incompatible architectures were manufactured for the entire vast USSR, for half of which at least basic software was hardly scraped off.
Even Soviet fanatics must admit that, compared to about 50 installations in the United States (with millions of lines of code), this was not just a failure, it was a disaster!
The list of problems facing Soviet informatics was repeatedly formulated based on the results of a heap of meetings, summarizing them, the following came out:
There is no single concept for the development of computing technology in the country. Each department produces the cars that it likes.
The country does not have a unified computer architecture, which means that there is no program portability.
The problem of portability is also very acute because the majority (about 50%) of programs are written in machine codes.
There are not enough programmers in the Union, just as there are not enough educational institutions that graduate them.
There is an acute shortage of application and system software.
The country does not have a unified computer architecture, which means that there is no program portability.
The problem of portability is also very acute because the majority (about 50%) of programs are written in machine codes.
There are not enough programmers in the Union, just as there are not enough educational institutions that graduate them.
There is an acute shortage of application and system software.
All these problems had to be urgently addressed.
It was absolutely unrealistic to develop from scratch in a couple of years a completely original architecture, to which IBM went 20 years (and to the development culture of which - another 50 years). The culture of the development of Soviet computers is well described in the memoirs of the programmer Samuil Lyubitsky:
Magnetic tapes were the most noteworthy feature of the Minsk-22 machine (the most emotionally noteworthy) and deserve a separate paragraph. They were very thick, very inelastic brown ribbons, four or even five inches wide. No bobbins. Can you imagine, no bobbins! How to work with them? That's how. There were two pockets in the ribbon duct, open at the top, like two incredibly deep glasses. It was necessary to grab the end of the tightly wound tape with your fingers and throw the skein into one pocket, then squeeze the end of the tape between the magnetic heads and build it into another pocket. In the process of movement, and the tape was used as a direct (!) Access device, it moved between the pockets, laying there in whimsical waves (the front wall of the pocket was glass, and you could admire it). And so that the tape did not slip entirely into one of the pockets, holes were made at the ends of it with an ordinary hole punch. In the tape conduit, there were special bulbs and light sensors, which were supposed (according to the designer's idea) to stop the tape at the last moment. Well, yes, of course! Every couple of hours in one of the conveyors, the tape slipped into a pocket. It was then that the fun began ... you need to stand on a special stool (since the top of your pocket is higher than your shoulders, and your hands should not get tired) and pick up a "fishing rod" - a flexible two-meter rod specially made of wire with a hook at the end. Well ... go ahead! We launch the fishing rod into the pocket. Remember, there is a hole at the end of the tape made by a hole punch. For this hole you need to hook the tape with a hook. Nonsense, you say? Yes, only a ribbon, it fits in her pocket as she pleases and you are very lucky if the end of it lies on top and you see it ... And be careful! The most valuable information on the tape is the results of a three-hour count. Scratch, damage the magnetic layer - ambit! Meanwhile, time is ticking, and the reports must be sent to the trust ... And you are standing on a stool with a fishing rod up and down, pull-pull, up and down, pull-pull ... Can you imagine what was said here about the unfortunate designer this miracle and its maternal relatives?
In general, this nightmare had to be stopped.
As for the programs, there can be no two opinions either. According to Doroditsyn, in the USSR in 1969 there were no more than 1 programmers, moreover, specialists in a heap of incompatible architectures, self-taught, mathematicians and physicists, and so on.
None of them were professionals, because the professional DEVELOPMENT of programs, and not writing as a hand went - we have not taught anywhere, and this is a separate complex discipline, which can easily be confirmed by any programmer. Brooks wrote (based on OS / 360 development):
… I suggest that you stick to the following rule: translators are three times more complex than conventional application programs, and operating systems are three times more complex than translators.
OS / 360 took, according to his estimate, 5 man-years; as a result, a project of comparable complexity would have been compiled by all Soviet programmers for 000 years at best. And that's not counting translators and thousands of applications. The famous odious passage of Babayan is known (which we will talk about separately in the part about "Elbrus"):
The calculation was that it would be possible to steal a lot of software - and computer technology would flourish. This, of course, did not happen. Because after everyone was herded into one place, creativity ended. Figuratively speaking, the brains began to dry from completely uncreative work. You just had to guess how Western, in reality outdated, computers were made. The advanced level was not known, they were not engaged in advanced developments, there was a hope that the software would flood ... It soon became clear that the software did not flood, the stolen pieces did not fit together, the programs did not work. Everything had to be rewritten, and what was taken out was ancient and did not work well. It was a deafening failure ...
Naturally, these are fairy tales.
Each question has two answers - pleasant and correct. Correct - usually painful. After the adoption of the EU, of course, the incredible dawn did not come, but the issue of software was resolved, in fact, before the collapse of the Union. The import programs finally worked fine without problems and dips and even without localization.
Given that 99% of the Union's technical progress was based on copying, there was no question where to get the cars, it is clear that from the United States. What exactly to copy was not a question either - obviously the best, S / 360.
Aside from the ideal niche of the mainframe line with millions of lines of software around the world, the S / 360 had several other important advantages.
First, she was going to GIS, which in the USSR had already been ripped off and mastered.
Secondly, its architecture was complex, at the limit of what the Union could copy (with Cray, as we remember, we could not cope with it), but not prohibitively complex. So, in fact, there was only one choice.
Which S / 360 implementation should I strip off - the original, UNIVAC 9000, RCA Spectra 70, English Electric System 4, or whatever?
On this occasion, there was a whole meeting, excerpts from it are widely known (its primary source is the famous book by BN Malinovsky "The History of Computing Technology in Persons"), and we will reproduce them. This conversation is covered in many places, but its interpretation, as a rule, is extremely one-sided. The same Malinovsky interprets it as follows:
From the discussion that took place, it is clear that Lebedev, Dorodnitsyn, Rakovsky, Sulim, Matkin were against copying the IBM-360 system; Keldysh said: "We need to buy a license and make our own cars, otherwise we will repeat what others have done." And Kalmykov hesitated - he listed the advantages of focusing on the ICL. The main active supporters of copying were the general designer of the ES COMPUTER Krutovskikh, his first deputy Levin, Shura-Bura, Przhiyalkovsky. If at a meeting with Kalmykov on December 18, 1969, where the final decision was made, the general designer spoke out against copying, computer technology in the USSR would have taken a different path ... Lebedev's unsuccessful attempt to change the decision was already mentioned. The refusal aggravated his illness, hastened the tragic denouement.
Apparently, this is where the roots of the myth grow about how Lebedev defended the original domestic developments.
In fact, things were a little different.
It is clearly evident from the conversation that the question - to copy or not - was not at all. There was a question, what to copy and the funny thing is that this question, in fact, was not there! Because, as we already said, English Electric System 4 is a clone of the RCA Spectra 70, a clone ... yes, the same S / 360! So Lebedev, Rameev, and all other patriarchs were unanimous on the issue of copying, only S / 360 will save Soviet informatics! The only thing they could not agree on was who to work with. With the GDR Germans from Robotron, who already pirated the original S / 360, or with the British from ICL, who offered to help set up the production of a clone - System 4.
So, this conversation was really epoch-making. That's just not entirely because of why they usually think. If you analyze it carefully, understanding the technical terms, you will see the following. There are two groups of academicians: one - for cloning a clone together with the British (conventionally: Lebedev-Rameeva and Deputy Minister Sulim, whom they convinced) and the second - for cloning the original together with the Germans (conventionally: Przhiyalkovsky - Shura-Bura). And the already known to us oak-headed and vindictive, like the devil, almighty Minister of the Radio Industry - Kalmykov, known for his love to bury designers who dared not to like him.
Naturally, Kalmykov, a former electrician-petroleum engineer, does not understand anything about the subject, Sulim understands better, after all, he, at least nominally, worked with Lebedev on the M-20. Therefore, Lebedev's group, using a variety of near-technical demagogy, is struggling to push work with ICL. Przhiyalkovsky's group, using no less demagogy, is trying to push through work with the Germans. This is clearly seen from the way they generally build a conversation, Kalmykov, on the other hand, is simply blinking, then talking nonsense. Other officials from the military-industrial complex and the Central Committee are just pieces of furniture, they understand the topic even less than Kalmykov.
Let's see the points.
So on December 18, 1969:
Present: Kalmykov, Keldysh, Gorshkov (chairman of the military-industrial complex. - Author's note), Savin, Kochetov (representatives of the CPSU Central Committee. - Author's note), Rakovsky (Deputy Chairman of the USSR State Planning Committee. - Author's note), Sulim, Lebedev, Krutovskikh , Gorshkov (Deputy Minister of Radio Industry. - Author's note), Levin, Shura-Bura, Ushakov, Arefieva, Przhiyalkovsky, Matkin, Dorodnitsyn.
Sulim... On the state of negotiations with the GDR and ICL.
Variant IBM-360. In the GDR, an orientation towards the IBM-360 was adopted. One of the models (R-40) is being successfully developed. We have a groundwork, we have a team that is able to start work. The development of the operating system IBM-360 will require 2200 man-years and 700 developers. There are no contacts with IBM. Difficulties will arise in acquiring an analog machine. Its cost is 4-5 million dollars. In the GDR, only part of the required documentation is available.
ICL variant. We will receive all the technical documentation, assistance in its development. We'll have to make some minor alterations. The firm offers to buy a batch of machines produced by it. It is possible to use a team of programmers to prepare application programs.
A group of our programmers are already undergoing an internship at the firm. In the future, joint development of a fourth-generation computer. The firm is trying to help in everything, as it hopes in alliance with European firms, including us, to compete with IBM. There is an agreement of Italian and French firms to participate in the creation of fourth-generation computing technology.
Sulim... On the state of negotiations with the GDR and ICL.
Variant IBM-360. In the GDR, an orientation towards the IBM-360 was adopted. One of the models (R-40) is being successfully developed. We have a groundwork, we have a team that is able to start work. The development of the operating system IBM-360 will require 2200 man-years and 700 developers. There are no contacts with IBM. Difficulties will arise in acquiring an analog machine. Its cost is 4-5 million dollars. In the GDR, only part of the required documentation is available.
ICL variant. We will receive all the technical documentation, assistance in its development. We'll have to make some minor alterations. The firm offers to buy a batch of machines produced by it. It is possible to use a team of programmers to prepare application programs.
A group of our programmers are already undergoing an internship at the firm. In the future, joint development of a fourth-generation computer. The firm is trying to help in everything, as it hopes in alliance with European firms, including us, to compete with IBM. There is an agreement of Italian and French firms to participate in the creation of fourth-generation computing technology.
Set out sensibly and to the point.
As we already wrote, the ICL was organized just a year before this meeting and immediately, seeing the once great computer industry in England completely dying down, rushed to establish contacts with the USSR.
Why with the Union?
Well, who else, not France, where at this time the development of computers had time to completely die. In addition, from 1964 to 1970, the Laborites were in power, traditionally looking towards socialism with sympathy and with obvious antipathy towards the United States, from under whose influence Britain unsuccessfully tried to crawl out for almost the entire XNUMXth century. The USSR was a natural and obvious ally in this. In addition, on the continent, the USSR was the only one with sufficient economic power and a potentially monstrously large market, and we had plenty of bright minds.
In addition, ICL offered fair partnerships. Training of our computer programmers and architects. Licensed clone. Improved microinstruction architecture compared to the original. Complete set of documentation. And yes, they wanted to make the next car together on an equal footing. It was a really great chance, and Lebedev, like Rameev, who were sincerely caring for the Soviet computer industry, understood this perfectly well.
The Germans, on the other hand, had nothing, except for jammed papers from S / 360 (and even then not all), which they, without help, enthusiastically on their own initiative, had been dismantling for more than a year at the ROBOTRON plant with the aim, like the Chinese, to collect a left copy for themselves and quietly bargain it in Europe in order to leave its competitor from Germany - Siemens, which had a license and officially sold clones.
They never had any bright plans to raise Soviet informatics from their knees. However, when they learned that the USSR was looking for a Western partner for the production of computers, then, of course, their eyes lit up, and how, for reasons similar to the British in terms of market size, we will master and master. It remains only to talk Kalmykov out. Przhiyalkovsky enters:
Przhiyalkovsky... For the IBM-360, we have a system of 6 thousand micro-commands, 90% of fuel and power plant diagrams, 70% routed, 7000 units of design documentation. When reorienting to ICL, this entire backlog will have to be reworked, this will delay the work by 1–1,5 years. It will take a lot of currency (for the purchase of ICL computers). The option of cooperation with the GDR, which is successfully working on the IBM-360, is preferable. If you strengthen the team of mathematicians, then the DOS can be developed by 1971. It is time to stop fluctuating.
Krutovsky... Our project envisioned a system of models IBM-360. When reorienting to ICL, the composition of the models should be different. Specifications are changing. It takes 4–5 months for a preliminary project. At ICL, there is no clarity on older models. They are added to a number of small and medium-sized computers as supercomputers. It is better not to do this. In case of reorientation, the terms of preparation of technical documentation will be delayed by 1,5-2 years, and maybe even more. Working with the GDR on the IBM-360, you can get DOS and OS by the beginning of mass production, the question of their development is removed. The Germans went further than us. They will not be able to reorient themselves. The British need a market. They will lead us by the nose. They will not cooperate on big cars. You can't buy 150 cars from them.
Krutovsky... Our project envisioned a system of models IBM-360. When reorienting to ICL, the composition of the models should be different. Specifications are changing. It takes 4–5 months for a preliminary project. At ICL, there is no clarity on older models. They are added to a number of small and medium-sized computers as supercomputers. It is better not to do this. In case of reorientation, the terms of preparation of technical documentation will be delayed by 1,5-2 years, and maybe even more. Working with the GDR on the IBM-360, you can get DOS and OS by the beginning of mass production, the question of their development is removed. The Germans went further than us. They will not be able to reorient themselves. The British need a market. They will lead us by the nose. They will not cooperate on big cars. You can't buy 150 cars from them.
Przhiyalkovsky and Krutovskikh were fully awarded for their idea, both were successively directors of NITSEVT (created from SKB-245 just for the EU series) and General Designers of ES EVM. A good career, especially for the Krutovskys, a man who has never created a single computer in his life has become a general, according to the best Soviet tradition (Przhiyalkovsky even worked on Minsk, as you can see, the designers of that very great Minsk were so excellently drowned for cloning). Krutovskikh understood perfectly well that in the case of working with the British, he and others like him would go to the forest, because ICL is interested in developers, and not promoted by the party. As a result, he says some kind of nonsense about a supercomputer, about the fact that the British will be led by the nose, that we will be 2 years behind (although with the Germans, who have “70% ready”, in the end they are four), that DOS in case Germans do not need to develop (sorry, ICL and so gave free and purely licensed ALL software for System 4, including OS ...) and so on.
Lebedevites enter again.
Dorodnitsyn... The question of mastering the IBM-360 is presented in a simplified form. Everything is much more complicated. It takes at least four years to master the OS, and it is not known what we will get. We must ourselves (together with ICL) create DOS and OS and focus on developing machines in conjunction with ICL.
Lebedev... The IBM-360 system is a series of computers that are ten years old. The number of machines we are building must be limited to machines of low and medium productivity. The architecture of the IBM-360 is not suitable for large models (supercomputers). The British want to compete with the Americans in the transition to fourth-generation computers. The higher the productivity of the machine, the more structural features it has. The British are laying down design automation. The software system for "Sistema-4" is dynamic; if there are contacts, it can be developed. This will facilitate the training of our own personnel. It is better to train them by developing their own system (together with the British).
Lebedev... The IBM-360 system is a series of computers that are ten years old. The number of machines we are building must be limited to machines of low and medium productivity. The architecture of the IBM-360 is not suitable for large models (supercomputers). The British want to compete with the Americans in the transition to fourth-generation computers. The higher the productivity of the machine, the more structural features it has. The British are laying down design automation. The software system for "Sistema-4" is dynamic; if there are contacts, it can be developed. This will facilitate the training of our own personnel. It is better to train them by developing their own system (together with the British).
Doroditsyn dryly comments that with the Germans we will find ourselves in a puddle (and this is what happened in the end).
Lebedev also begins to carry heresy, most likely to speak at the intellectual level of Kalmykov, with arguments that he understands. Naturally, the S / 360, which by that time had only been on sale for 2 years, could not “become outdated by 10 years,” here Lebedev is lying and does not blush. Well, he is also disingenuous about his favorite supercomputers, the S / 360 model 95 could break even the CDC6600.
Everything else is absolutely true: S / 360 is damn complicated, and at times it is better to make a copy of it with the British (and continue to work with them on the next generation), they have experience, design tools, they are ready to teach and share.
Supporters of the Germans are taking the floor again.
Shura-Bura... From the point of view of the software system, the American version is preferable. The OS needs to be improved. To do this, you need to know all the programs.
Keldysh... You need to buy licenses and make your own cars. Otherwise, we'll just repeat what others have done. Basically, you have to build big cars yourself.
Keldysh... You need to buy licenses and make your own cars. Otherwise, we'll just repeat what others have done. Basically, you have to build big cars yourself.
Shura-Bura with all due respect to him, it seems, is not at all in the subject.
He was a mathematician and programmer (the same academic, from an ivory tower) and vaguely represented the subject of the controversy. He was told that the Yankees had more programs - he believed, although all the same programs were running on System 4. Alas, Shura-Bura grew up in the Soviet zoo and, apparently, the idea that different names of cars = different software, firmly stuck in his head.
Keldysh shows why, of all those present, he is the head of the USSR Academy of Sciences. He answers like a real politician - so as to do, but not to do, to copy, but to develop his own, under a license, but it is not clear who and whom. In general, the talent of the demagogue as it is.
Whose side is he on?
Yes, on his own, who has what to copy - he does not care, he conducts speeches so that both parties consider him theirs.
Lebedev... Our mathematicians believe that it is better to train programmers according to the methodology of the British.
Rakovsky... You need to think about perspective. A unified concept is needed. Everyone said that IBM's software was better, but the OS was cumbersome. It cannot be fully mastered within four to five years. It’s difficult, but today we need to make a decision. If you focus on the ICL, it will be difficult with the GDR; in five years, the Germans will produce 200 copies of the P-40. Still, the ICL proposal should be accepted.
Krutovsky... All developers, except for Rameev, do not want to reorient themselves to ICL. The P-50 will be ready in 1971.
Kalmykov... The presence of DOS immediately makes it possible to use the machines that we will start producing. We can get a lot of programs from the Germans. Negative points. We don't have IBM-360 machines. And we will not have contacts with IBM. If we reorient ourselves to ICL, we will waste time. But with them direct contact and cooperation is possible in the creation of a fourth generation computer. This is a great advantage. They will do the fourth generation of computers without the Americans, they want to be competitive with IBM.
Rakovsky... You need to think about perspective. A unified concept is needed. Everyone said that IBM's software was better, but the OS was cumbersome. It cannot be fully mastered within four to five years. It’s difficult, but today we need to make a decision. If you focus on the ICL, it will be difficult with the GDR; in five years, the Germans will produce 200 copies of the P-40. Still, the ICL proposal should be accepted.
Krutovsky... All developers, except for Rameev, do not want to reorient themselves to ICL. The P-50 will be ready in 1971.
Kalmykov... The presence of DOS immediately makes it possible to use the machines that we will start producing. We can get a lot of programs from the Germans. Negative points. We don't have IBM-360 machines. And we will not have contacts with IBM. If we reorient ourselves to ICL, we will waste time. But with them direct contact and cooperation is possible in the creation of a fourth generation computer. This is a great advantage. They will do the fourth generation of computers without the Americans, they want to be competitive with IBM.
Deputy Chairman of the State Planning Committee of the USSR Rakovsky demonstrates his "deep" knowledge of the OS / 360 architecture, then complains that well, how can we throw the Germans, they tried for us here! But nevertheless he makes a choice for the British.
Krutovskikh again said how he cut it off - one of your Rameyev is stirring up the water, the rest have been all for a long time. Kalmykov hesitates.
And then Keldysh unexpectedly finishes:
Keldysh... You should not reorient yourself to ICL, but negotiations with them on the fourth generation of computers should be conducted.
Kalmykov... We will not reorient to ICL. Let the Germans ask the question of helping more. "
Kalmykov... We will not reorient to ICL. Let the Germans ask the question of helping more. "
Well, in general, that's all, it happened.
Lebedev's party could not push through their point of view, after that Sulim and Rameev really demonstratively put the papers on the table and left their posts, not wanting to see what would begin next, and Lebedev really got sick with grief, and Kalmykov can indirectly count the third ditched constructor.
As a result, an evil fate seemed to hang over the USSR.
Our competitors, for obvious reason, could not develop their competitor from scratch. Copying is, in principle, not such a bad option, AMD, for example, was founded as a direct clone of Intel, and still has been releasing a common architecture for them for 50 years and does not think about dying.
At the same time, the culture of developing computers was completely absent in the USSR, and it would not have been possible to simply take and successfully copy the S / 360. But then, lo and behold, the ICL firm falls on its head, with which you can at least try to do everything right. Their experience and technology, our money and intellectual resources - not the fact that it would have worked, but the attempt would have been worthwhile.
And now, standing on the threshold, one must, in the best Soviet tradition, stumble over this threshold and again kiss his head! So it was with literally everything that the USSR undertook in terms of developing a computer, indeed, some kind of curse.
How many times have we started something good - with the machines of Kartsev, Yuditsky (three times each!), Our own microprocessors and mini-computers, an attempt to develop a copy of the CDC 1604 for science, an attempt to shake the IBM throne together with the British.
And every time everything came down to the elementary.
The Soviet system, in principle, did not have any checks and balances, literally a couple of narrow-minded, greedy, limited and vindictive people at the top could ditch everything with a few words. At the same time, ironically, only such people, as a rule, went upstairs and ended up. Hence the eternal cognitive dissonance that arises in many when reading the history of domestic computers, it's just some kind of continuous bummer for 40 years.
Naturally, the purely Soviet EUs did not take off in the form in which they were intended.
Despite the Germans, it was possible to master production only by 1971 (junior models), indeed 5 years behind and only increasing this gap. The quality of the self-assembly of such complex equipment without the help of the British was terrifying - according to the recollections of many, the first series of "Series 1" that they met were remembered for the fact that they did not work at all, and it took months to debug them. Documentation was completely absent, people on the ground somehow, swearing, dealt with the channels to which they connected the disgusting Soviet periphery.
In general, the EU is remembered by most of the people as something monstrous, an anomaly that should not have been.
Here is a typical recollection of a person who worked with Soviet PCs:
... in the 90s I worked in the same office as a programmer on Iskra-1030, this was such an analogue of the IBM PC XT. One fine morning, the machine, when turned on, completely refuses to show anything on the display. An electronics engineer is called by phone, who, while holding a screwdriver around the victim, diagnoses "replacement of the video card". Well, and accordingly, leaves for a new one. Comes back after some time with a similar new board, externally - exactly the same. We change. We turn on the computer. He gives an error in the POST area with a digital code and then refuses to boot completely. We climb into the printed manual for the unit, into the list of errors during loading. What I saw there, I will remember for the rest of my life - opposite the error code given to us by the computer, a short decryption was written: NON-RYAZAN ASSEMBLY.
And it was a pitiful "Iskra", but imagine how, with this level of culture, it looked like assembling a car 50 times more difficult ...
After that, it is not surprising that most of the EU of a purely Soviet assembly (everyone dreamed of getting a GDR one) in half of the cases was installed in an inoperable state and was finished by the forces of local staff. It is not surprising that in parallel with the EU, Minsk-32 was also produced for a long time, and BESM-6 was discontinued altogether, count only in Russia.
Equally, it is not surprising that all serious firms continued to joyfully rivet their zoo, as well as work with "Minsk", MIR and BESM-6 until the mid-late 1970s, until the childhood diseases of the EU were cured, and comfortable and powerful clones already S / 370 on ECL-chips of the 500th series.
At the same time, as we said, they created as many as 2 generations of Elbrus, slowly sawed a clone of Cray-1 "Electronics SS BIS", gave birth to clones of the first PCs in the throes, and under the EU brand they developed a huge bunch of independent experimental machines - the Yerevan matrix special processor EC2700 , the Kiev EC2701 macro pipeline, the Leningrad multiprocessor with the dynamic architecture EC2704, the Taganrog multiprocessor EC2706, the family of multiprocessor systems PS-1000 / PS-2000 IPU of the USSR Academy of Sciences, the Kronos station and other amazing things, each of which must be discussed separately.
But then the economy of the USSR made a somersault in an ever-accelerating dive, and from the mid-1980s it was no longer up to computers.
The most unfortunate series of Soviet cars, not to blame for this. ES EVM. Remote control from EU 1022-02 (https://www.computer-museum.ru)
In this case, neither the architecture itself, nor even the idea of cloning is to blame.
The only fault is the crooked Soviet implementation, which (not a fact, but quite possible!) Could have been much better, because there was nowhere else to be worse.
Nevertheless, the EU, released in the amount of more than 15, nevertheless satisfied the Soviet computer hunger a little, and their tons of software helped the USSR to hold out right up to 000.
In general, a huge number of myths are associated with the EU series, almost more than with BESM-6, including incredible legends that Minsk-32 was more powerful, IBM stole the idea of firmware from the Soviet MIR computer (and in general, it was the first personal "computer, IBM was so impressed that she bought it right at the exhibition, although there is generally a murky story, it is impossible to find anything even about the exhibition itself in Western sources, let alone a purchase, and the only source of information about this very fact of purchase is Malinovsky's book without any references at all) and so on.
In general, one can talk about it for a long time, only one thing is obvious - this is both the most glorious (due to the fact that it was thanks to the EU that the computerization of the USSR finally came true), and the most tragic (because of how it came true) part of our history.
Now, having examined the rise and fall of the main part of Lebedev's school, it remains for us to highlight the last hero of Soviet informatics, who had the most direct relation to the missile defense project, and at this end the cycle.
Further on the program - Burtsev and the incredible story of his "Elbrus".
Information