A computer reservation system offers 79 million fares and 6 million schedules. It processes 1.2 million fare changes each day. Sabre alone processes 48 million transactions a day, including all shopping, pricing, and and availability queries.

ComputerWorld offers a glimpse into the challenges that Sabre faces in handling all of this computing at a reasonable cost.

In order to rein in escalating processing costs and offer customers more options, Sabre is completely overhauling the software used by airlines, travel agents and passengers to find and book flights. In stages, Sabre is replacing its old mainframe assembler code with modern languages running on cheap commodity computers and open systems, including Linux.

…

Three forces hugely increased the processing demands on Sabre systems over the years. After airfares were deregulated in 1979, travel agents began shopping based on both price and schedules, not just schedules. Then, in the late 1980s, travel agents began using PC-based automated search tools that continuously scanned Sabre databases for the lowest fares. Finally, in the mid-1990s, consumers on the Internet joined travel professionals in shopping for flights. Sabre’s processing economics took an ugly turn as the “look-to-book” ratio soared. Looking for the best schedules and fares generates data-processing costs but no revenue; it isn’t until someone books a flight that anyone makes money.

…

But, Murphy says, while the look-to-book ratio skyrocketed and pricing and scheduling options proliferated, three technological forces came to the rescue: Moore’s Law, open systems and ubiquitous standards. Moore’s Law — which states that the amount of computing power available per dollar doubles every 18-24 months—enabled Sabre to assemble a scalable farm of powerful servers built around cheap commodity processors and huge memories.

…

Murphy says the cost of the project will exceed $100 million, but he won’t be more specific. Results so far have been encouraging. “We sold the project on the basis of reducing total cost of ownership by 40%,” he says, “I actually think we’ll surpass that. Running a query—Dallas to Chicago, say—on the new system is about 80% cheaper.”



And, he says, developers are getting 100% productivity gains because they are working in higher-level languages—Java and C++—and because the application architecture is so much easier to debug, change and enhance than the old mainframe assembler code.



Finally, airlines are getting the ability to put in new options and features quickly and cheaply. Last year, Sabre announced SabreSonic, a suite of services that enable airlines to tap into Sabre systems and databases in order to offer passengers new services, such as streamlined airport check-in.



“Changes are much easier to do, and much less expensive,” says Gianni Marostica, president of Sabre Airline Solutions. “This allows us to implement things airlines think of on the fly.”

“We have to be in an environment where our cost, two years from now, is half what it is today on a per-unit basis,” Healy says. “The air-shopping problem will be more complex, there will be more Internet users and there will be more [users] internationally. The only way we can meet that demand is by riding Moore’s Law.”