big machines | Locklin on science

The largest computer ever built

Posted in big machines by Scott Locklin on March 28, 2013

While cold war jets are an old interest of mine, almost everything built to fight the cold war fascinates me. All ages are characterized by madness; only a few have that madness captured in physical objects. Consider the largest computer ever built: the “Semi-Automatic Ground Environment” or SAGE system.

The SAGE system was designed to solve a data fusion problem. Radar installations across North America kept watch against Soviet bombers. These needed to be networked together and coordinated with air defense missiles and interceptors. Seems simple, right? In the 1950s and 1960s, this was not simple. The country is big; hundreds of radar stations and sensors needed to be integrated. It wasn’t as easy as it was in England in WW-2, when enemy aircraft location was plotted by hand on maps as the radar data came in: North America is much larger, and the planes traveled much faster in the 50s and 60s. No group of people could really make the decisions in time to mount an effective defense. You needed some kind of computer to make the decisions.

These giant electric brains took up an acre or so of real estate, and were encased in huge windowless concrete pillboxes all over the country.

StewartAFBNY_SAGE-DC2_Life-8

The SAGE system had many firsts: it was the first nation wide networked computer system. While it used special leased telephone lines and some of the first modems (at a blistering 1300 baud), it was effectively the internet, long before the internet. It was the first to use CRT screens. The first to use a “touch screen interface” via the use of light pens on the CRT. It was the first to use magnetic core memory. It was the first real time, high availability computer system. It was the first computer system to use time sharing. Many people attribute the genesis of computer programming as a profession to the SAGE system. Modern air traffic control, and computer booking systems of course, descend from the SAGE system.

StewartAFBNY_SAGE-DC2-Scope_Life-7

Each of the 27 computers that made up the system was a dual core 32 bit CPU made of 60,000 vacuum tubes 175,000 diodes, and 12,000 of those newfangled transistors. Memory: 256k of magnetic core RAM (invented for this project) clocking in every 6 microseconds. These things weighed 300 tons, consumed 3 megawatts of electrical power and ran a blistering 75,000 operations per second. The dual cores weren’t used for multiprocessing. One was kept on hot standby in case the other failed. Since our grandfathers knew about fault tolerance, they had a system to replace the tubes before they burned out: downtime was typically only a couple hours a year.

sage_bb14

Each one drove 50 to 150 GUI workstations, and interacted with more than a hundred radars, interceptors and missile batteries. Remember that the next time you whine that your computer ain’t fast enough. No. This thing is less powerful than even a shitty cell phone (the 386 was probably approximately equivalent), and it did significantly more than your PC does. Each one was capable of coordinating the air defense of the entire North American continent.

sage_bb13

It is worth pointing out that these machines not only ran all that equipment, and dealt with all that data, they also guided interceptors to their target locations. The F-106 and F-102 could be directly controlled by the SAGE system after takeoff. We think of “drones” as the latest and greatest newfangled thing in warfare: they have actually existed for a very long time. In many ways the SAGE system was more impressive than, say, the Predator system.

SAGE_floorplan_large

Another interesting piece of the SAGE system was the BOMARC missile system. The BOMARC (made by Boeing and the Michigan Aeronautic Research Center -which no longer exists) was primarily ramjet powered, and carried either a small nuke, or a half ton of conventional explosives. It was entirely dependent on the SAGE system for guidance to target. It was also incredibly stupid and dangerous: the original rocket boosters used hypergolic fuels, and would occasionally spectacularly explode in their silos, spreading dangerous plutonium around.

The SAGE system started running in 1958, and didn’t stop until 1984. Was it necessary? Like many interesting cold war artifacts, SAGE was more or less made obsolete by missiles around the time it deployed. While it did cost around $90 billion in 2013 dollars, it also was responsible for a good fraction of the technological things we now take for granted. Only barbarians do not remember their history, so anyone involved in modern technological projects should study it for lessons in engineering practice on long term and large scale projects.

First lesson I take from the SAGE system: solving the right problem. SAGE solved an important problem, that of air defense from enemy aircraft. It did so beautifully. The problem was, by the time it was deployed, bomber attack was a secondary issue: the primary threat was ballistic missiles. The US probably needed something like it anyway, but it is worth noticing that long-term, large-scale projects could very well be made redundant by deploy time.

Second lesson I take from the SAGE system: assemble a team that knows how to solve similar problems. MIT already had a computer on hand which constituted half of the solution, so it made perfect sense to scale up some parts of MIT into the MITRE corporation and Lincoln labs. Let’s say the government got a bug up its ass to spend $100 billion building a computer with the same capabilities as a dog’s brain, or one that could program itself. What institution would be most qualified to do this? I can’t answer this question, because pretty much nobody knows how to do something similar.

The third and final lesson I’ll take from the development of SAGE: break down the problem into manageable pieces, and solve them. They used the technology on hand in the 50s; vacuum tubes, telephone lines and CRTs. They didn’t postulate any significant breakthroughs in order to get ‘er done. They made do with what they knew was possible As such, the path to success was obvious. Engineering genius came along the way. If you don’t have manageable pieces, you don’t have a real project: you have a wish. What are the manageable pieces needed to make “nanotech” or controlled nuclear fusion a reality? What are the manageable pieces needed to make quantum computing or deriving all electrical power from the sun a reality? I don’t know, and I don’t know of anybody else who does: therefore, such things do not count as legitimate long term projects.

“One of the outstanding things… was the esprit de corps—the spirit that pervaded the operation. Everyone had a sense of purpose—a sense of doing something important. People felt the pressure and had the desire to solve the air defense problem, although there was often disagreement as to how to achieve that end. Energy was directed more toward solving individual problems, such as making a workable high-speed memory or a useable data link, than it was toward solving the problem of the value of the finished product. It was an engineer’s dream.” -John F. Jacobs Former Senior Vice President The MITRE Corporation

Insanely fascinating SAGE manuals found here:

http://bitsavers.trailing-edge.com/pdf/ibm/sage/

49 comments

The Mars Science Laboratory is a Rube Goldberg contraption

Posted in big machines by Scott Locklin on August 4, 2012

Mars missions have a troubled history; lots of outright failures have happened -more than half if you count international missions. Why NASA thought it was a good idea to take this many risks on a high-dollar mission … well, I dunno. Maybe it isn’t as risky as it looks, but I wouldn’t want to have to make book on the success of this mission. All kinds of things can go wrong; from the supersonic parachute, to the radar, to the weird maneuvers (how about … just cutting the damn parachute?), to the wacky “sky-crane” idea (um, what about using shrink-wrap, dudes?). I hope it works, and will cheer for them if it does, but I’m not real optimistic.

That said, the rover is a really cool thing.

Edit add: Success!

Nothing like hindsight bias to make something look easy!

2 comments

The B-58 Hustler

Posted in big machines by Scott Locklin on June 2, 2012

Aerospace technology became grotesque and beautiful in the 1950s. One of the most grotesque and beautiful creations of that bizarre era of technological excellence was the Convair B-58 Hustler.

As I have said before, the early supersonic era is one of my favorite epochs of aerospace technology. Not because the products of this era were particularly effective: most of them were spectacular failures. I love this era of technology because it is beautiful. The creations of this time stretched technologies to the absolute limits. It is the embodiment of a tremendous strain of the human technological spirit, grasping after the unattainable ideal. It is the technological manifestation of the look on the face of the Olympic weight lifter, as his joints crack and sinews pop, straining every muscle fiber to accomplish the impossible goal.

What was the B-58? Convair more or less took its F-102 airframe, multiplied its linear dimensions by 1.5 or so, strapped four enormous engines to the wings, stuck a machine gun on its tail, and called it a Supersonic Bomber. The engines were so fuel hungry, the only practical way to get the thing to Russia was to integrate the bomb with the external fuel tank, but it worked. It was bristling with new technologies. In addition to its ability to travel at ludicrous speeds, the B-58 would actually talk to its pilots in a lady-like voice which they called sexy-Sally (or “the bitch” since it always delivered bad news). Yeah, your dumb iphone will also do this, but imagine what some corn-pone who used to pilot a big aluminum dump truck like the B-17 made of this! It must have been considered something like magic.

Other innovative technologies: the airframe used a sophisticated aluminum honeycomb and fiberglass composite material. The ejection seats were wacky clamshell devices to make it possible for all 3 occupants of the aircraft to eject safely. It had several sophisticated radars and avionics that included an accurate inertial guidance system that used a sort of electric star sextant, doppler radar, and electric compass to reset the drift (probably utilizing some analog electro-mechanical Kalman filter like ideas, which is pretty neat if you’re a signal processing nerd).

The combat doctrine of the era was that nuclear weapon equipped bombers would fly very high, and go very fast, to avoid enemy interceptors. This, effectively, was the combat doctrine of WW-2, applied to the supersonic era. This actually made sense for a brief period: it was very difficult to shoot down a B-29 in late WW-2 which flew too high for most interceptors. In order for interceptors to shoot down the B-29, they would have to gain altitude very quickly, and be able to beat the speed of the B-29, which was considerable. Remember, the B-58 was flying only 12 years after the introduction of the B-29. It was conceived a mere 5 years after the deployment of the B-29. What happened 5 years ago, technologically speaking, in modern aerospace technology? 5 years ago is pretty much last week. I think 5 years ago, we were saying the F-35 would fly in 5 years. Which is pretty much what we’re saying today. But the 40s and 50s were a different era. People were more serious, and technological advance was happening so quickly, combat doctrine couldn’t keep up.

What was the effect of the B-58? Well, for one thing, it freaked out the Soviets. The built a whole new generation of ridiculously fast supersonic interceptors to deal with it. They also built their own supersonic bombers, like the Tu-22 ”Blinder” pictured below. They built a giant radar network, so they could at least see the things coming in time to send something up to get the B-58. They also invested heavily in surface to air missile technology, which eventually made the B-58 irrelevant.

The B-58 was rendered irrelevant by surface to air missiles almost as soon as it was deployed, but it earned a reprieve by virtue of being reasonably good at flying low, under the SAM radar. Even so, it was preposterously expensive to purchase and operate. It was estimated to cost its weight in gold. The subsonic B-52, conceived years before, was almost as good at flying under the radar as the Hustler was, and it cost a lot less, carried more boom and had a much larger combat radius. Truth be told, the altitude and cruising speed of a B-52 wasn’t all that different from that of a B-58 either; the Hustler could only maintain supersonic flight while burning tremendous amounts of fuel. While the supersonic dash capability of the B-58 might have made it a more survivable platform in the event of a nuclear war … since a nuclear war with the Soviets would have been the end of civilization, it really doesn’t matter much if the B-58 survives, does it? What’s it going to do when it gets back? In an uncharacteristic display of intelligence, Robert McNamera decided to phase out the Hustler in 1965, a mere 5 years after it was first deployed, and was completely retired by 1970. I don’t think he did it for the right reasons, as McNamera was one of the most cement-headed SecDefs we have ever had, but at least he did it.

What can be learned from the experience of the B-58? I think it is clear by now: we don’t need supersonic bombers. We don’t need supersonic anything to drop bombs. Supersonic flight doesn’t help much in this role, and the compromises the capability inflicts are not worth the trouble or the money. The other supersonic-capable bomber we deployed, the B-1B, was never particularly useful either, other than as a nuclear codpiece for freaking out Russian people. The only way a very fast, high flying bomber could conceivably be useful is if it goes very high, and very fast, at which point it’s basically an insanely expensive, pointlessly reusable ICBM anyway.

The other thing we can learn from this: yesterday’s combat doctrine isn’t very helpful. We’re presently engaged in developing a next generation stealth bomber, as if the last generation were not expensive enough. While I admit the utility of a stealth bomber for strategic targets, or the early parts of a war on a third world country, I’m not sure why we need more than we already have. Finally, absurdly expensive weapons systems are almost never worth the money. While we’ve gone through several generations of “advanced” bombers since the Hustler, when we need to drop lots of bombs, the venerable B-52, a 1948 design, is still the tool of choice.