By Rusty Weston

Mark Depathy, Senior Infrastructure Engineer at People’s United Bank, a 165-year-old institution based in Bridgeport, Conn., is serious about creating systems that can handle the bank’s mergers and acquisitions. He’s also serious about meeting new, increasingly stringent business and regulatory demands for tighter security and governance. That’s why Depathy and his IT colleagues are also serious about mainframes.

“One of the mainframe’s strengths is scalability,” Depathy says. “You can add new applications without expending a lot of new resources. You don’t need additional floor space. In most cases you can integrate backup and recovery of these new apps into existing infrastructure.”

Depathy and People’s United Bank are far from alone. The mainframe, having survived its midlife crisis, is making a comeback. CIOs are discovering that mainframe systems are secure, economical, powerful—and surprisingly energy-efficient.

These benefits are driving new and growing demand for mainframes, and giving CIOs new reasons to consider — or reconsider — an older technology. Of course, IBM’s current zSeries hardware is hardly your father’s mainframe. IBM, now the sole maker of mainframe hardware, has recently invested billions of dollars to make its zSeries mainframes run faster, consume less energy and — still a work in progress — become easier to program and manage, too.

Similarly, solution providers that cater to the mainframe market are offering new services for CIOs. For example, CA offers consulting services through its Mainframe Value Program. Under this program, CA consultants work with CIOs by first taking an inventory of their mainframe software; second, selecting a top few products to carefully investigate; and third, presenting findings and suggesting options for greater value, including insights into best practices. “CA is making a commitment over the next five years to build best-in-class mainframe support, maintenance and optimization capabilities,” says Mark Combs, Senior VP and General Manager at CA. “Support isn’t just about code; it’s also having knowledgeable people offering best-in-class advice.”

This advice comes to a mainframe marketplace that is pulling itself out from roughly a decade’s worth of downturn. In the late 1990s and early 2000s, the mainframe was hurt by hardware rivals offering commodity- priced x86 and Linux/Unix servers. Many CIOs chafed at rising mainframe costs, too. But then, in response to this cross-platform competition, IBM dramatically improved the price/performance ratio of its mainframe hardware system. According to market watchers at Gartner, the current cost of mainframe MIPS—millions of instructions per second, the most common measure of mainframe performance — is now two-thirds of what it was in 2001. That marks an average price/performance improvement of nearly 10 percent a year.

All this trying harder has paid off. IBM says its mainframe business is now healthy, with four consecutive quarters of revenue growth and seven consecutive quarters of growth in MIPS usage by customers (see chart, below).

The Mainframe Advantage
CIOs reconsidering the mainframe need to be aware of three important developments in the marketplace. First, today’s mainframe is no longer an island. Instead, big iron is increasingly well connected with small desktops, not to mention Web sites, third-party suppliers and customers— often with service-oriented architecture (SOA) applications that give new life to older legacy applications.

Second, better price/performance, coupled with wider use of server virtualization, has helped drive consolidation of multiple servers onto fewer — yet more powerful — platforms. For example, IBM made news in 2007 by consolidating nearly 4,000 distributed servers in six locations onto just 30 Linux mainframes, saving $250 million in the process.

Third is greater energy efficiency. Numerous tests have shown big iron to be more energy-efficient, MIPS for MIPS, than distributed servers. In fact, the more a company uses mainframes, the greater its energy savings versus PCs or rack-mounted servers. (See related story, “Greening the Data Center”)

That’s not to say the mainframe is poised to supplant all distributed servers. Even with these successes, deployments of distributed servers over the past decade have outpaced those for mainframes. “The growth today is more centered on volume servers,” contends Mark Monroe, Director of Sustainable Computing at Sun Microsystems, which makes servers. “Today’s most compute-intensive applications use low-cost, open-system servers.”

Nonetheless, the resurgence of CIO interest in the mainframe is real and, analysts say, sustainable, if for no other reason but the high cost of moving. “Having to live with legacy apps means that the mainframe platform will be sustainable for quite some time,” says Rakesh Kumar, a Research VP with Gartner. “To move those apps to a different platform would be tremendously difficult.”

If this decade has taught CIOs anything, it’s that the mainframe’s success isn’t a zero-sum gain. Even though other platforms may grow faster, for certain workloads mainframes remain the best possible tool. Many banks and other financial services business, for example, are keeping legacy, mission-critical applications on the mainframe, even as they offload other work to PC and midrange servers. “CIOs are starting to say that the promises of the distributed platform have not been realized on the same scale as the mainframe,” says Christopher O’Malley, a SVP at CA and General Manager of its mainframe division. “That’s especially true for applications that support global market demand and therefore need to be up and running all the time.”

At People’s United Bank, the mainframe’s intricate operational procedures, painstakingly documented to fulfill regulatory requirements such as Sarbanes- Oxley, have spurred executives to appreciate the platform’s value over smaller alternatives. For example, the bank uses a tape-encryption solution that runs on the mainframe. “You read where a truckload of tapes is lost, stolen or misplaced,” says Depathy. “If that happens [here], it’s unreadable by anybody but us.”

Help Wanted
The growth of mainframe MIPS in the enterprise raises one persistent concern on the part of CIOs: not the state of technology, but the people. Or, rather, the lack of college graduates prepared to manage and operate mainframes today. “There’s a shortage of mainframe administrators and programmers who are willing to learn older-generation languages such as Cobol,” maintains John Abbott, Chief Analyst and Research Director at The 451 Group, an IT analysis firm. “That’s likely to get worse over time, though more open-systems management tools should help a little.”

The mainframe talent shortage is a problem that has crept up on CIOs slowly, but can’t be turned around quickly, experts say. “Most people with my skills have gray hair or no hair,” concedes Depathy of People’s United Bank. O’Malley of CA adds: “If you were to look at a bell curve of people working on mainframes, you’d see the average age is 55. It’s not an issue now, but it will be in 10 years.”

Of course, if the average is 55 years, then many mainframe vets can be expected to hang up their flowcharts soon. “By 2012 a lot of the people with these skill sets will be retiring,” asserts William DiBella, President of AFCOM, an association of data center professionals. To avoid a true talent crunch, DiBella hopes that today’s seasoned data center operators will mentor a younger generation of workers on such mainframe skills as tape management, maintaining legacy Cobol applications, and working in the Assembly language.

But to reach this younger generation, providers of mainframe systems and solutions may also need to adapt. “Mainframes aren’t point-and-click yet,” says Depathy of People’s United Bank. “So one of the things that scares people away is complexity.”

Depathy notes that both IBM and CA are among the vendors attempting to simplify mainframe applications. He hopes these changes will make mainframe systems easier to understand and, ultimately, boost their appeal with younger workers. In fact, IBM says it will spend $100 million over the next five years to make its mainframes easier to program and manage, even over the Web. Big Blue also promises new tools that will allow non-techies to program mainframes.

Reducing operational complexity in the mainframe environment is a good idea that is coming to pass, slowly. Most of the complexity experienced by customers is tied to software in one form or another. For its part, CA intends to help reform software licensing practices by instituting simpler contracting processes and terms. For example, the company recently introduced CA IDMS Server r16.1, a solution offering capabilities for secure, auditable extension of mainframe data and business logic across the Web and SOA environments. CA also recently announced “Day One” support for the z/OS 1.9 operating system across all its mainframe solutions.

This and other new solutions have also disabused skeptics of the notion that CIOs must choose between the mainframe’s monolithic, centralized approach and a decentralized, distributed model with midrange servers, blades or PC servers clustered in a variety of locations. Thanks to the emergence of Web services, mainframes and servers are likely not only to co-survive, but also to collaborate better in the future.

Looking ahead, Gartner analyst Kumar predicts that we may be nearing the end of the zSeries cycle. He expects to see a new generation of IBM mainframes arriving in 12 to 18 months. If the recent past is a guide to what’s to come, reports of the mainframe’s death will be seen as greatly exaggerated.

Rusty Weston is a San Francisco-based journalist who blogs for FastCompany magazine and MyGlobalCareer.com.