SAGE is also the name of a system administrators' professional organization.
SAGE Sector Control Room. The screen shows the US eastern seaboard from the extreme west end of Nova Scotia in the upper right, to Philadelphia in the lower left. Cape Cod is centered, and two "targets" are being tracked offshore. All images courtesy MITRE.
SAGE, the Semi Automatic Ground Environment, was an automated control system for collecting, tracking and intercepting enemy bomber aircraft used by NORAD from the late 1950s into the 1980s. By the time it was fully operational the Soviet bomber threat had been replaced by the Soviet missile threat, for which SAGE was entirely inadequate.
Nevertheless, SAGE was tremendously important; it led to huge advances in online systems and interactive computing, real-time computing, and data communications using modems. It is generally considered to be one of the most advanced and successful large computer systems ever developed.
IBM's role in SAGE (the design and manufacture of the AN/FSQ-7 computer, a vacuum tube computer with ferrite core memory based on the Whirlwind) was an important factor leading to IBM's domination of the computer industry.
Prior to the introduction of SAGE, the task of intercepting bombers was becoming increasingly difficult. In World War II radar had pushed the advantage in favour of the defender, detecting a raid at long range and thus giving defenders plenty of time to launch their interceptor aircraft. The RAF could leave their fighters on the ground until the raid was detected, guiding them over radio towards the bombers, calculating directions "by hand".
In the post-war era, jet-powered aircraft decreased the detection time, but also the time needed to intercept the bombers. But while the speed of the aircraft increased, the time taken to direct a particular interceptor to a particular bomber remained largely constant. This included tasks such as collecting information about the targets, figuring out where they were going (developing a track), deciding what planes should intercept them, telling everyone, and then tracking both the interceptors and the bombers to an interception point.
A study in the 1950s by the RCAF concluded that it would take on the order of one minute per interception. With flight times on the order of an hour by several hundred aircraft, some were bound to escape interception due to operator overload. With nuclear bombs onboard, this was unacceptable. The problem became even more acute if the bombers attacked at low level. Radar is line-of-sight, so by approaching close to the ground they would remain hidden behind the curvature of the Earth until approaching to within a few tens of miles (tens of km). With a jet bomber, this meant the defenders had only a few minutes to react, far too little time to launch an interceptor.
It was this problem that particularly bothered Dr. George Valley, an MIT physics professor. In order to provide any sort of protection for the entire US, a series of radar stations would have to span both coasts and across Canada. In the event of a raid, there would simply be far too many reports to be able to successfully guide interception. His solution was automation, connecting all of the radar sites to a computer which would then control all of the incoming and outgoing flow of information. The interception operator's workload would be greatly reduced; they simply had to tell the computer which targets to attack, and perhaps choose what assets to use. All of the communications would be handled by the computer, and would be effectively instantaneous.
This would require the system to update the operators in real time, and the only system in the world capable of doing this in 1948 when Valley studied the problem was the Project Whirlwind computer at MIT. The Whirlwind project, originally intended to control a US Navy flight simulator to train bomber crews, had run into problems and the Navy was losing interest. Valley talked to Jay Forrester, leader of the Whirlwind project, and together they wrote a study proposal to use Whirlwind for air defense.
The US Air Force was interested, and in 1949 they provided funding under the name Project Charles to develop a demonstration system. Information from several radars in the Cape Cod area was forwarded to the Whirlwind, which then developed tracks for the targets being reported. The project was a qualified success, and the Air Force took over the project under Project Claude, moving development to the new MIT Lincoln Laboratory in 1954. Making a military-grade version of the Whirlwind was a massive project that required close connections between Lincoln Labs, industrial partners who would build the machines and communications, and the military. In order to provide oversight and management during the deployment phase, MITRE was formed in 1958 to take over the project.
Production of the resulting machines, known technically as the AN/FSQ-7 but almost always referred to as the Whirlwind II, was initially awarded to RCA but later given to IBM, who started production in 1958. The buildings and internal power supply and communications were provided by Western Electric, phone lines by the Bell System, and the software, 500,000 lines of assembler, by a spin-off of RAND Corporation called SDC.
The AN/FSQ-7 used 55,000 vacuum tubes, about 1/2 acre (2,000 m²) of floor space, weighted 275 tons and used up to three megawatts of power. Although the failure rate of an individual tube was low due to efforts in quality control, so many were used that the daily failure rate was in the hundreds. Each center had staff dedicated to replacing dead tubes by running up and down the racks of machinery with shopping carts filled with replacements. The AN/FSQ-7s remain the largest computers ever built, and will likely hold that record in the future. Each SAGE site included two computers for redundancy, with one processor on "hot standby" at all times. In spite of the poor reliability of the tubes, this dual-processor design made for remarkably high overall system uptime. 99% availability was not unusual.
SAGE operator's terminal. The operator is selecting a target with the light gun. Not visible is the built-in ash tray.
SAGE sites were connected to a number of tracking stations which sent in sighting reports over a teletype system connected over normal telephone lines. Reports were typed in by operators in a specific format, which the SAGE computers then collected and displayed on a CRT as icons. Operators at the center could select any of the "targets" on the display with a light gun, and then display additional information about the contact reported by the tracking stations. Up to 150 operators could be supported from each center.
When a target turned out to be interesting, SAGE also helped him select a proper response. Reports similar to those from the radar sites kept the SAGE system up to date with information on the availability and status of various weapons and aircraft, including all airfields, BOMARC and Nike Hercules anti-aircraft missile sites. When the operator chose one of these to intercept the target, orders would automatically be sent via teletype to local controllers who would take over from there. Additional messages would also be sent to higher headquarters, as well as other SAGE centers.
A massive building program started along with continued work on the computer systems and communications, with the first groundbreaking at McChord AFB in 1957. The buildings were huge above-ground concrete bricks that were often placed near cities without the residents being aware of what they were. The first SAGE Division became operational in Syracuse, New York in January 1959, and by 1963 the system was already complete with 22 Sector Direction Centers and three similar Combat Centers. When NORAD was set up another site added in North Bay, Ontario in Canada, although in this case the entire SAGE system was buried deep underground in what became known as "the hole".
The total engineering effort for SAGE was immense. Total project cost remains unknown, but estimates place it between 8 and 12 billion 1964 dollars, more than the Manhattan Project that developed the nuclear bomb SAGE defended against.
Typical SAGE site. On the far left are cooling towers for the generators located in the (low) middle building. The "cube" has four floors, with air conditioning and wiring on the ground, the computers on the second floor, offices on the third and the combat center on top. The "big screen" shown at the top of this article extends from the third to fourth floor.
The SAGE system was operational until 1979, when it was replaced by newer systems and airborne control. However, the North Bay system ran until 1983 when it was dismantled and sent to The Computer Museum in Boston. In 1996 the remainder was moved to Moffett Federal Airfield for storage and is now in the collection of the Computer History Museum in Mountain View, California.
Questions about the ability of the SAGE system to actually handle a "hot war" situation were continuous. On one occasion SAC was able to penetrate the defenses, and on other occasions huge flocks of seabirds were tracked as a potential bomber attack. A more serious problem was that by the time the system was fully operational, the USSR had already started deploying ICBMs, making SAGE largely useless.
SAGE was, for all intents, an air-traffic control system. This was not lost on the FAA, who used SAGE systems in their own automated control systems, many of which remained in service until recently. The system also gave IBM valuable insight, and it was not long after that the CEO of American Airlines met one of the IBM people involved in SAGE by accident on a flight, and soon the two companies were developing the SABRE airline reservation system.
Other major SAGE developments included:
- CRT-based real-time user interface
- use of wide-area communications via modems
- John F. Jacobs, The SAGE Air Defense System: A Personal History (MITRE Corporation, 1986)
- Robert R. Everett (editor), Special Issue: SAGE (Semi-Automatic Ground Environment), Annals of the History of Computing, Vol. 5 (No. 4), 1983