The mainframe's legendary reliability and availability can't be attributed solely to state-of-the-art technologies. It's a cultural mind-set that grew up around the data centre, says Robert DiAngelo, CIO at MIB Group. "The disciplines needed to manage data processing are very well defined and controlled," he says.

Alan Walker, vice president at Sabre Holdings, agrees. "We've spent years building a culture around [the mainframe]," he says.

In contrast, the culture around distributed systems, which grew up from departmental computing initiatives within individual business units, has been much less disciplined. Modernisation efforts that don't take culture into account are destined to fail. "Some of the biggest failures in IT history have been associated with package migration. You must change the culture, not just the software," says Gartner analyst Dale Vecchio.

Sabre had to address that before rewriting its fare-search application and migrating it to an open-systems environment. "When you deal with tens of thousands of transactions per second, you can't reboot," Walker says. A staff that manages open systems doesn't always understand the best practices required to maintain that kind of uptime.

A successful migration to open systems can't happen unless those mainframe values migrate as well, Walker says. "Outages aren't caused by the operating system or hardware," he says. They're produced when the programmer or operations staff does something wrong. "Open-systems [staff] may run a little looser," but veteran mainframe programmers and staff have been indoctrinated not to make those errors, Walker adds.

The key is to infuse that culture into the entire staff. That starts by bringing the open-systems staff into the mainframe world - literally. "I'm taking my guys up there, doing bonding with the veteran mainframe guys," Walker says. DiAngelo's mainframe staffers are also learning Java and doing the migration on the mainframe. Robert Rosen, president of the IBM mainframe user group Share, sees a trend of organisations hiring experienced mainframe staffers to run distributed data centres. "They realise that they need that kind of discipline," he says.