Charles Fitzgerald, Microsoft’s general manager of platform strategies, insists that he has no axe to grind about big iron. But over the past year, he says, "we’ve had a whole bunch of customers come to us and say, ‘What can you do to help us get off the mainframe?’”
Fitzgerald says that these pleas derive partly from the fact that the mainframe guys are retiring and partly from IT’s intense focus on lowering costs. “What people have figured out is they’re paying obscene quantities of money. A Wintel server today is about £1 a MIP. A mainframe MIP is about £1,000. That’s three orders of magnitude.”
That bang-for-the-buck argument makes sense to Bob Offutt, senior vice president and chief architect at Sabre Holdings, a $2 billion company that handles approximately 40 percent of the world’s travel reservations. At Sabre, a million new reservations is considered a light day and peak hours see about 14,000 transactions per second. On a quest to lower operating costs, Sabre is four years into a migration from mainframes to Intel-based servers from Hewlett-Packard.
Offutt’s mainframes run TPF (Transaction Processing Facility), a highly reliable legacy operating system designed for large transaction volumes. Until recently, all programming was done in an assembler, making the very notion of flexibility a joke. Yet the rise of the Internet, with Sabre powering such high-traffic sites as Travelocity, has piled on new business demands.
“Shopping for airfares has gone through the roof,” Offutt says. “Where the look-to-book ratio used to be on the order of 8-to-1 or 10-to-1, it now exceeds 30-to-1. So we need a lot more shopping capacity.”
Offutt has already migrated the shopping and pricing functions for U.S. domestic air travel to a Linux server farm running MySQL, taking the cost savings argument a bit further than Microsoft may like.
Offutt says that by the end of next year, he hopes to be off the legacy systems completely. Already, he has seen a significant drop in development costs simply because he can use programmers writing in modern languages.
He estimates that the new infrastructure will reduce the cost of each transaction by 80 percent at the end of next year, when he plans to have fully migrated booking and fulfilment as well.
Phenomenal mainframe reliability underlies the “If it works, why fix it?” argument that has kept big iron chugging along for years. But when demand leaps ahead of current capacity - and cost-effective horsepower is the solution - the time may be ripe to put the old warhorses out to pasture.