Tails Through Time: Reservisor

Showing posts with label Reservisor. Show all posts

16 June 2015

American Automates Reservations and Ticketing, Part Two: SABRE

SABRE terminals at American's terminal at JFK Airport

Part One of the story of how American Airlines moved to automate the passenger reservations and ticketing process looked how things were handled manually until the development with the Teleregister Corporation of the Reservisor System which was subsequently upgraded to a more capable system called Magnetronic Reservisor. Charles Amman, an analyst with American Airlines who was tasked by chairman C.R. Smith to improve the process, had delineated that there were three steps to making an airline reservation- the first step was to determine seat availability for a given flight, the second step was to then adjust the seat inventory to that a given flight didn't get oversold, and the last step in the process was to connect the passenger's name and information to that reservation to issue a plane ticket. Magnetronic Reservisor was one of the earliest distributed data processing systems to try and automate as much of that three step process as possible. The system could perform the first two tasks, but it took a second system developed by IBM called Reserwriter to complete the third step. Between the second step in Reservisor and the third step in Reserwriter, there was still a need for manual intervention and this was time-consuming. In 1958, it could take Magnetronic Reservisor up to three hours to complete that three step process outlined by Charles Amman! Things weren't going to get easier not just on account of the growth of passenger traffic on American, but the impending 1959 inauguration of Boeing 707 jet services. The speed of jet aircraft would, quite simply, outstrip the speed at which the Reservisor/Reserwriter system could process reservations and issue tickets.What C.R. Smith wanted was an automated system that could perform all the three step process in near-real time.

How that was solved, though, requires us to take a step back a few years from the year 1958. In October 1952, the United States Air Force and IBM initiated the development of SAGE (Semi-Automatic Ground Environment), an integrated set of radar, surface to air missile batteries, and interceptor aircraft all tied together by digital computers and long distance telephone lines. At the time of the inception of SAGE, there was only one digital computer in existence, the Whirlwind at MIT. Very quickly, the SAGE project would eclipse the Manhattan Project in terms of size and complexity as much of the technology had to be developed to develop the world's first integrated air defense system. IBM would win the contract in April 1953 to develop the computers that would be the core of SAGE. As SAGE would tie together all the assets of the joint US-Canada North American Air Defense Command (NORAD), it was absolutely necessary for SAGE to be able to update and process data in real time so air defense units could respond rapidly to any threats. It was also necessary that SAGE be reliable.

The IBM 7090 mainframe was developed from the SAGE FSQ-7

Called Whirlwind II by IBM but AN/FSQ-7 by the USAF, the FSQ-7 was the world's first production digital computer weighing 275 tons with 919 miles of cable, 50,000 vacuum tubes, 7,300 pluggable units and 170,000 diodes. The computer consumed 3 megawatts of power and required 100 operators. The FSQ-7 was the brains of SAGE. The magnetic core of the FSQ-7 had 256 kilobytes of memory and was supported by 150 kilobyte magnetic drum as well as 4 tape drives each holding 100 kilobytes of data. One-hundred fifty consoles displayed data on circular CRT screens that used light pens to interact with the screen information. With the light pens, operators could tag radar tracks and issue commands to air defense units. Each of the 24 SAGE director centers built across the United States had a large multistory hardened concrete structure housing two FSQ-7 computers- one running and the other one operating as back up to the first. Each of the director centers were then linked not only to each other but to also as many as 100 air defense units by long distance telephone lines. SAGE was the single most expensive defense project until the development of the ICBM. The first direction center went online at McGuire AFB in New Jersey in November 1956. Today we'd call computers like the FSQ-7 a mainframe computer and in fact, the FSQ-7 is the progenitor of all modern mainframe computers used today. Much of what IBM learned in the development and implementation of SAGE led to the development of their first business mainframe computers, the IBM 7090 being an development of the FSQ-7. But at the time, IBM had spent much of their time working with the US military and were lagging behind their competitors in getting real-time data processing technology for commercial and business applications.

One day on an American Airlines flight in 1953, two men sitting next to each other struck up a conversation that sprang from a realization they shared the same surname, Smith. One man was Blair Smith, an IBM sales executive who was bemoaning IBM's late entry into the business world and how much he needed to make a sale. The other man was none other than American's chairman, C.R. Smith, who bemoaned the monumental challenge the airline faced with the limitations of Reservisor with the inevitable arrival of faster jet aircraft. When the flight landed at New York La Guardia, Smith invited the IBM executive to tour the Reservisor nerve center located at the airport. It wasn't long before American and IBM entered into a partnership for a new, more advanced system that would be much more capable than Reservisor/Reserwriter. American's specifications for the new system were demanding given the technology of the day:

Be able to maintain at least 40,000 passenger reservations in real-time (this would be approximately 9 times what Reservisor was capable of handling).
Be able to support 83,000 phone calls each day (Reservisor could handle only 15% of that load).
Be able to retrieve flight information and seat availability within 3 seconds 90 percent of the time.
Be able to support remote operations with at least 100 locations nationwide.
Be able to record and access passenger information (what Reserwriter was doing already but it was a separate system).
To have the capacity in the future to send and receive messages to other airlines.

American reservations agents at SABRE terminals

Nothing of this capability had ever been done in the business world before and this was where IBM's experience with the USAF in developing SAGE paid off. SAGE was the only existing distributed network computer mainframe system that could do data processing in real time. Partnership with American was just the break into the business world IBM needed for its mainframe computer systems. After his fortuitous meeting with C.R. Smith, he informed the head of IBM, Thomas Watson, who astutely realized many new technologies that IBM was developing based on its SAGE experience were just what would meet American Airlines' stringent specifications. IBM code-named the project SABER (Semi-Automatic Business Environment Response). Discussions between IBM and American continued for approximately two years under a memorandum of understanding that defined SABER as a joint project. American Airlines even set up its own department of software programmers to work with IBM to create the operating system to be used by SABER. On 18 September 1957, IBM submitted the formalized SABER proposal to C.R. Smith and the board of American Airlines which quickly approved the joint venture. Smith saw seat inventory and rapid booking and ticketing as not just an important business function but also both a customer service function and a marketing advantage for American- many other airlines were working on getting similar systems to SABER, but not all of them were as advanced and as scalable as SABER. In the same year that the Boeing 707 inaugural flights for American Airlines were flown in 1959, the project was expanded to include future upgrades as technology allowed, one of which was the ability to request seats on other airlines and to be scalable and support upgrades that would increase speed to near real time. That same year, one of Smith's subordinates came across an ad in a magazine for the Buick LeSabre and suggested that the project name be changed slightly to SABRE as it had more marketing panache.

Advertisement showing how the SABRE network worked.

The first two IBM 7090 mainframe computers were installed in a data center in Westchester County north of New York City in 1960 on a trial basis and the system was an immediate success. The SABRE terminal had a display that agents would slide coded cards into that would then query the mainframe on possible flight options. A series of buttons on the left of the card allowed individual flights to be queried on seat availability and a modified IBM electric typewriter (that became the iconic IBM Selectric typewriter as a stand alone unit) allowed the agent to enter passenger information into SABRE much like what IBM's earlier system, Reserwriter, did for the airline. SABRE terminals were quickly installed throughout American's network, from city ticketing offices to airport check-in counters. By 1964, SABRE had completely taken over all the reservations and ticketing functions from the Reservisor system. With 10,400 dedicated long distance telephone lines for SABRE, the system became not just the largest single user of telephone lines in the United States but the largest civilian data processing system in the entire world when it became fully operational in 1964.

Typical SABRE terminal

The Transport Workers Union wasn't happy with SABRE, though. The union was concerned that jobs would be lost with the automatic functions of the system and C.R. Smith promised that not one employee would lose his or her job in the transition from Reservisor to SABRE. This was perhaps a bit coy on Smith's part as the reservation agent force at American had the highest turnover rate in the airline so it was just a simple matter of not replacing normal losses via attrition. Switching to SABRE would end up saving American a very significant 30% in labor costs overall. But this wasn't the only area American won out, there was obviously the competitive advantage as SABRE was considerably faster than competitor airline systems which gave American the edge in customer service. Flights no longer had to be undersold and there would come to be a greater degree of utilization of American's fleet as it was SABRE generated reams of data on passenger loads, peak flights, seasonal variations and more. Even more significant, it allowed the airline to keep track of its most loyal customers.

But as much as an advantage as SABRE was for American Airlines, all that data the system generated from thousands of passenger bookings, fare transactions, was accumulating faster on tape reels in SABRE facilities than American's own people could review it. Data processing is one thing; data analysis, on the other hand, is using that data to optimize the business and in the mid-1960s, that was still an emerging field in the airline industry. For the first few years SABRE was operational, American was content to sit on its own laurels, but it wasn't long before IBM turned its SABRE experience into a new business as it extended the technology to other airlines. It wasn't long before American's competitors had systems of their own with IBM mainframes. Rather unusually given American's history of innovation in the industry, SABRE wasn't given the same priority by C.R. Smith's successors at the helm of the airline. That would change soon enough when American hired an aggressive, astonishingly detail-oriented former TWA executive as its new chief financial officer in 1973.

That man was Robert Crandall. His background and passion for detail would turn SABRE into a competitive weapon that would make American Airlines one of the world's most dominant airlines for a good twenty years. But as always, that's a blog posting for another day here at TAILS THROUGH TIME!

01 June 2015

American Automates Reservations and Ticketing, Part One: Reservisor System

C.R. Smith knew reservation systems had to keep up with jet speeds

In the early days of the airlines, the reservations and ticketing process was all done by hand at a central office with a ledger that was passed among agents who recorded reservations and erased ones that were canceled. As airlines grew in the early 1930s with the introduction of larger aircraft like the Boeing 247 and in particular, the Douglas DC-3, a single book was obviously impractical so multiple ledgers for each flight was kept in a large lazy-Susan with the reservations agents sitting all around. As they took incoming calls, they retrieved the ledger for that particular flight. The books had to be frequently reconciled so that more seats weren't sold than available and even for small airlines, this process of inventory reconciliation was time consuming on top of the fact that it had to be done multiple times through the day. Books soon gave way to large chalk boards and reservation offices had to get bigger in size to not only accommodate every larger boards to display the seat inventory for each flight, but to also accommodate the increasing numbers of personnel to keep the system running. As agents took calls, they filled out cards that were then taken clerks to the other agents who maintained the boards. Supervisors watched the boards and in the case for large airlines, they would have to use binoculars or opera glasses to view the data. A typical reservations office for a large airline might be sixty agents taking phone calls with forty additional clerks and agents who maintained the boards and conveyed data back and forth. It was a personnel intensive and time consuming process become progressively more cumbersome as airlines added more flights and larger aircraft. The first systems in place were called "Request and Reply"- agents took calls and then passed a message to "Inventory Control" who were the agents who maintained the boards. The message went back to confirm that the seat was available and reserved and a third message confirmed the reservation. So even the most rudimentary system in use needed three messages and even in those days, the airlines were the biggest users of telephone lines in the United States. In 1939, the Boston reservations office for American Airlines had determined that until a flight was 80% sold, it was possible to sell seats freely without having to get a reply from inventory control. This streamlined the process as message volumes were cut in half. When a flight was 80% full, agents in inventory control sent out a stop-sale and any further sales on those flights required confirmation from inventory control. This system was called "Sell and Report" but became "Request and Reply" for any flight over 80% booked. Despite this, the airline still had to reconcile seat inventory as well as get out ticketing information to the airports. Clerical mistakes were common and it wasn't unusual for business traveler to book two flights just in case.

Typical reservations center with status boards (this one is Pan Am's)

In 1943 a system analyst with a background in electronics and communications, Charles Amman, joined American Airlines and was tasked to find a better solution. Running the military's Air Transport Command during the war, C.R. Smith astutely saw not just a postwar boom in air travel but the arrival of faster and larger aircraft that would hopelessly encumber the manual system. Amman had determined there were three key steps in the reservations process: determine seat availability, adjust the seat inventory, and record the passenger's name. Amman met with several manufacturers of adding machines and electromechanical equipment of the day and every single one thought what American wanted was impossible- too many variables to adjust in real time was simply beyond what the technology of the day could accomplish. Amman even pared down the requirements to just the first two steps in the reservations process, determining seat availability and adjusting the seat inventory but even this was too daunting for companies he approached. Amman had C.R. Smith's support to try and come up with an in-house solution. A series of tall cylinders with each representing a flight on a given day were filled with marbles. Each marble represented an available seat. An agent would use a cash register like device to select a flight to make a reservation. This sent an electrical signal to the cylinder that represented that flight and a hatch dropped one marble out of the cylinder. If a reservation was canceled, a marble was electrically released back into the cylinder. adjusting the seat inventory automatically. Amman's machine was bulky and impractical for an airline the size of American, but it demonstrated the principles of a solution.

An early Reservisor terminal

Amman's mock up convinced the Teleregister Corporation to sign on the develop the reservations system for American. Originally a division of Western Union, Teleregister was set up in 1948 as separate company and produced the stock ticker units used by stock brokers. This gave them the experience in remote data transmission that could be applied to the American reservations project. American called the system "Reservisor" and the first pilot system was a massive room of switches, relays, and plugs called "The Brain" which was installed in the Boston reservations office in 1946. Electrical relays replaced the marbles and cylinders of Amman's mockup with the agents using a smaller device that looked like the early adding machines used by accountants. "The Brain" replied on the agent's terminal if space was available- a green light meant space was available for booking, an amber light meant the flight was sold out. A one year pilot of the Reservisor at the Boston office showed two hundred additional passengers could be booked each day with twenty less personnel. C.R. Smith, back in the executive suite after serving as head of the Air Transport Command during the war, was excited about the possibilities of expanding the Reservisor system across American's network. But the Reservisor as it was still encountered delays as it only solved the first two parts of the reservations process- determining seat availability and adjusting seat inventory. Passenger names for the tickets were still taken down by hand and there was still a stop-sale system in place as agents were responsible for manually inserting control plugs into the proper parts of "The Brain" when a flight reached 80% full. But the Reservisor as it was was a quantum leap over what American had done before and because the airline wanted accuracy and reliability, Teleregister incorporated redundancy into the Reservisor system. The challenge at that point wasn't just addressing the last step of the reservations process but to also speed up the system and increase its memory capacity as the airline not only grew but added faster aircraft like jets. Recognizing that there was a sales advantage to a faster system, Amman also wanted an upgrade to Reservisor that would allow an agent to recommend an alternate flight should a passenger's first choice of flight be booked. What he wanted was for Reservisor to be able to not just store increasing amounts of flight information, but to also be quick to allow agents to easily query the system to find out seat availability on alternate flights. He had been following developments in computer memory and settled on what was called drum memory for Reservisor.

Magnetronic Reservisor. Note the smaller terminal and drum memory units.

The drum was a large metal cylinder coated with a ferromagnetic material and a row of fixed read-write heads along the long axis of the drum. Each read-write head had its own track on the drum. It was a precursor to the hard disk platters of today which have a moving read-write head that moves across a spinning disk. In drum memory, each track on the circumference of the drum had its own read-write head. Drum memory is what replaced the relays and switches of "The Brain" of the first version of Reservisor. The new upgraded Reservisor was called the Magnetronic Reservisor and first went into operation in 1952 at American's reservations office at New York La Guardia Airport. The system had a response time of one second and the drum unit could hold data for 1,000 flights up to 10 days in advance. Four years later, a larger unit was installed in the West Side of New York City that had a half-second response time and could store data for 2,000 flights up to 31 days in advance. Amman redesigned the West Side reservations center of American Airlines to take advantage of the new Magnetronic Reservisor with 362 agents to take phone calls, each with their own terminal. Forty additional agents were devoted just to handling reservation requests from travel agencies and large business accounts. Another 140 agents acted as liaisons between the reservations center and the other reservations offices, airport stations, and airline operations center. Forty supervisors oversaw the West Side center that averaged 4,500 phone calls a day.

Close up of the drum memory units. The wires ran to the read-write heads.

Despite the advances of the 1956 version of the Magnetronic Reservisor, Amman worked with IBM on the last step in the reservations process, connecting passenger information with the reservation. In 1956, the Reserwriter went into operation after testing in Buffalo. The Reserwriter basically read keypunched cards that represented passenger information and then automatically sent messages via teletype to the main reservations offices. Messages could go back and forth between the two systems (albeit manually) and for the first time for the airline, agents didn't have to be in the main reservation offices to book flights. By 1958, Reserwriter terminals were in key locations across the country in American's network. Despite the success of the whole system, it still suffered with errors on account of the number of individuals still required to make the system work. Approximately 8% of reservation transactions at American were in error despite the advances in the system. Passengers who benefited the most from the Reservisor system were those flying into and out of the largest of American's destinations. The Reserwriter terminals at outlying locations helped, but because manual action was needed to move information between Reserwriter and Reservisor, it was prone to error not to mention still time consuming. For example, a round trip reservation between Buffalo and New York La Guardia required 12 people, 15 distinct steps and could take as long as three hours. Passenger growth at American meant that over an eight year period from 1950 to 1958, the number of passengers flying on American per reservations employee had dropped from 5,100 to 3,100. Simply adding employees wasn't a solution in C.R. Smith's eyes. The transaction error rate meant the airline had to undersell a flight to avoid overselling a flight during times of peak volume. American would be taking delivery of its first Boeing 707s in 1959. This meant that a transcontinental flight was faster and could move more passengers- the airline's flagship DC-7 services moved 80 passengers across the United States in 10 hours but the 707s promised to move 112 passengers across the nation in just 5 to 6 hours. Smith was worried that the new jets would overwhelm the existing system. What American needed to keep up with the jet age was reservations system that moved just as fast, preferably in real-time and with as little manual intervention as possible. Charles Amman's original reservations problem from 1953 was still an issue that lurked in the background- determine seat availability, adjust the seat inventory, and record the passenger's name. Only now, it had to be done in real time if it was to keep up with jet aircraft.

That solution will be the subject of a future post on this blog, so stay tuned!

Source: Hard Landing: The Epic Contest for Power and Profits That Plunged the Airlines into Chaos by Thomas Petzinger. Times Business/Random House, 1996, pp 57-60. Waves of Change: Business Evolution Through Information Technology by James L. McKenney, Duncan C. Copeland, et all. Harvard Business Press, 1995, pp 97-106. Images: Wikipedia, American Airlines C.R. Smith Museum