Big Blue's mainframe gathers no rust

Mainframe customers are taking a fresh look at the Big Iron that celebrated its 40th birthday last month. IBM is spurring things along with new pricing schemes; more powerful processors; support for non-proprietary technologies such as TCP/IP, Linux and Java; and on-demand offerings that put the mainframe in the reach of even the smallest customers.

Hewitt Associates' mainframe environment is a classic example of how the Big Iron has evolved. Daniel Kaberon, director of computer resource management at the human resources outsourcing and consulting firm in Lincolnshire, Ill., says the company still uses the mainframe as the foundation of its data center. But it has integrated the system into an evolving architecture that links the mainframe with grid computing and blade servers to get more processing power at a lower price.

Nevertheless, some industry observers still see the mainframe as a dying breed.

"You can change and adapt the environment in which [the mainframe] works, but you can't turn it into something which it is not," says Bob Djurdjevic, president of Annex Research. "People will have to get rid [of the mainframe] eventually, just as they got rid of key-punch machines and a myriad of other systems."

Djurdjevic says the decline in mainframe revenue illustrates the trend. He says IBM has seen revenue drop - save for an uptick last year, the first upswing in eight years - to where it accounts for about 4% of Big Blue's overall revenue. But that's still around $4 billion, according to IDC estimates.

IBM doesn't break out revenue for specific product lines, but did note in its annual report that zSeries revenue was up 7.4% in 2003, in large part because of the release of the z990 "T-Rex" last spring. In the first quarter this year, IBM reported that total delivery of zSeries computing power as measured in million instructions per second nearly doubled compared to the same quarter a year ago.

"This doesn't come out and tell me that the mainframe is a dying platform," Kaberon says.

He still runs his firm's most critical applications on a Parallel Sysplex cluster of eight mainframes. But for those applications - or parts of applications - where the mainframe doesn't make sense, Kaberon isn't shy about making modifications.

Last fall, Hewitt moved a calculation engine that figures pension benefits off the mainframe and onto a grid of Linux-based, two-processor 2.8-GHz blades from IBM. He says costs associated with each calculation have dropped by more than 90% because it no longer uses expensive mainframe processing power.

Today, Kaberon is working on expanding the grid to support a composed print application that he wants to move off the mainframe, as well.

"We're using the grid as the back end of the mainframe," says Kaberon, who will lead a session focusing on this mainframe-grid architecture later this month at the Grid Today '04 conference in Philadelphia. "We really look at the grid as a mainframe compute peripheral. It's a co-processor. You put the application on the mainframe and then spit part of it out to process and then bring the results back in. The whole workload is managed from the mainframe."