This is from a case study I wrote in 1996:
These days we hear about technological change all the time. Sometimes, however, we hear about a technological change that may affect an entire industry. That’s the case of a new application developed by SABRE Decision Technologies, a division of American Airlines, for Air France.
Think about the many flights every day and the many different fares on all of those flights. Until recently, airlines were limited as to how they could maximize their capacity and profits by the ability of mainframe computers to crunch huge amounts of data. A new application, called the Availability Processor, makes it possible to enhance capacity and profits by dynamically calculating seat availability on fares for any combination of connections and destinations in real time. This application fits into the existing airline-reservation system and expands the current seat/fare availability calculations.
The possibilities for efficiency and increased profitability are impressive. The development method and speed with which the Availability Processor was developed are impressive as well. I spoke to Richard Green, Principal at SABRE Decision Technologies, about the development of the Availability Processor.
The goal for the Availability Processor was to develop a more efficient, profitable method for filling airplane seats and work with the other existing systems used for airline reservations. The previous availability system was based on a mainframe using flat files. It was relatively slow and unable to manage the data in a way that allowed real-time availability calculations based on revenue. The old system, for example, would take a whole night for the system to update static revenue control information. A real-time system was not even considered.
The speed with which VERSANT can handle complex data made the possibility of real-time availability calculations possible. What sort of speed are we talking about? First, take the size of an airline-reservations system into account. There are several hundred thousand airline-reservations terminals tapped into the system with more than one million transactions a day. The time-out factor for a real-time availability calculation is 750 msec. The Availability Processor easily meets that.
The business need driving this is the airlines’ need to maximize revenue on the seats they sell. The mainframe systems used static data that failed to take connections into account when pricing seats. Some seats were under-priced. Also, the mainframe systems did not have access to customer profile. The Availability Processor is positioned to significantly increase revenue for airlines using this additional data. At the same time, it will take many fewer people to maintain the Availability Processor.
More on the general topic: Myth: ODBMS Products Do Not Scale