The Mainframe at Forty: The Future Never Looked So Bright

The long-term relevance of the mainframe has been one of IBM’s major themes. If 2003 was any indication, there's no doubt that customers agree. As the mainframe prepares to celebrate its 40th birthday this April, analysts take note of its success, and explore how organizations are planning to put the increased processing power to work.

The long-term relevance of the mainframe has been one of IBM Corp.’s major themes over the last decade, even in the face of often-contradictory market research and sales data. If 2003 was any indication, however, it appears that reports of the mainframe’s death have been greatly exaggerated. In fact, as the mainframe prepares to celebrate its 40th birthday next month, some folks argue that its future has never looked so bright.

Want proof? Consider IBM’s strong showing in the enterprise server market during the fourth quarter of 2003, which was driven by a sharp spike in zSeries sales. According to market researcher International Data Corp. (IDC), Big Blue garnered 37.9 percent of worldwide server market revenues, up 17.7 percent over the year-ago quarter. IDC found that IBM received a big boost from surging sales of zSeries systems, which were up 33 percent over 2003.

It’s a safe bet that the new z990 (“T-Rex”) mainframe systems Big Blue unveiled in May of last year had something to do with this surge. IBM billed its new z990 as the largest and most powerful mainframe system ever built, and Bill Zeitler, senior vice president and group executive of IBM’s Systems Group, said that Big Blue had invested more than $1 billion in a T-Rex development project that spanned four years and involved the efforts of 1,200 engineers. For many current and potential customers, this amounted to an explicit vote of confidence for a platform that some industry watchers argued was consigned to irrelevance.

“It demonstrated a tremendous commitment to the mainframe,” says David Mastrobattista, zSeries marketing manager with IBM. “Usually when we announce a mainframe, when we’re talking about a new generation, it’s sometimes one-and-a-half to two times the capacity the prior of the generation. With the z990, we took that all the way to three times the capacity of the prior generation. It was unprecedented.”

Although demand for T-Rex was slow to ramp up (in the system’s first few months of availability, Big Blue trumpeted only a few customer wins), by early 2004, T-Rex was on a roll. In mid-February, for example, Big Blue notched a mega-deal with Sparkassen Informatik, a German company that provides software and IT services for nearly one quarter of Germany’s banks and financial institutions. Sparkassen Informatik planned to tap T-Rex’s unprecedented capacity to support a massive consolidation project: Moving from 32 z900 mainframes to 20 z990 systems.

“The new application they wanted to use for the zSeries was designed to support [their] core business over the next four years and to grow based on that, and to do this in the most cost effective manner there was a project defined to consolidate their workloads” on to the 20 z990 systems, says Michael Brander, an IBM representative who’s attached to the Sparkassen Informatik account.

Big Blue’s z900 and z800 mainframe systems appear to have plenty of life left on them as well. A week after touting the Sparkassen Informatik deal, IBM announced that Postbank—a German financial institution that serves 11.5 million customers—planned to replace its existing Fujitsu-Siemens mainframe infrastructure with new zSeries and pSeries systems. The German banking giant purchased four z900 mainframes for use as SAP database servers, IBM said.

IBM’s Mastrobattista says sales of the z990 grew pretty much as IBM expected. “We had a very steady ramp-up. I think part of that was the fact that we announced the z990 in May, but then we accelerated some features coming on to the platform, such as our secure key encryption,” he observes.

Another important contributor to zSeries sales, claims Mastrobattista, was the new “Mainframe Charter” that Big Blue unveiled in July. The idea, he says, was to affirm IBM’s commitment to Big Iron and provide a framework that organizations could use to plan future investments in mainframe-related technologies. “This was an attempt to show the marketplace IBM’s commitment to the platform, and that the zSeries platform was going to remain a strategic platform in the IBM eServer line.”

The ISV Connection

Of course, one would hardly expect IBM to downplay the prospects of its most lucrative platform, but Mastrobattista’s optimism is echoed by several independent software vendors; some sound positively bullish on Big Iron for the first time in years.

“We’re starting to see our customers adopt and implement the z990 …, and we know of a couple of cases where customers have never run a mainframe [but] are now expressing interest [in doing so],” confirms Jonathan Adams, director of research and development for BMC Software Corp.’s MainView product line, who notes that BMC had one of its most profitable quarters in recent memory, thanks in part to healthy sales of MainView, which were up 26 percent over Q3 2003, and up 131 percent sequentially over Q2 2004: “We’re measured as a publicly-traded company, primarily by revenue, whether it’s new licensing or maintenance revenue, and MainView had a nice December quarter, as did BMC.”

What it all adds up to, industry watchers say, is that customers aren’t quite as eager to move mission-critical applications and data off of the mainframe.

“The mentality now seems to be ‘get the best from your existing assets,’ so there’s no longer the debate about replacing host systems with Unix, or client-server, or whatever,” says Peter Havart-Simkin, senior vice-president of strategic development with host integration specialist NetManage Inc. “Organizations have learned that there really aren’t too many boxes out there that can get close to the reliability [of the mainframe].”

As a result, argues Eric Varness, director of product management with host integration specialist WRQ, host access and host integration technologies have assumed new importance. “Companies aren’t abandoning these [mainframe] systems. In fact, in some cases, they’re purchasing new ones. So there’s starting to emerge this concept of Web-to-host or legacy rejuvenation where the customer wants to extend this [mainframe data and applications] to non-traditional users, and they’re realizing that character-based [terminal emulation] just isn’t going to work,” he comments.

Naturally, you’d expect both vendors to make sky’s-the-limit arguments about their bread-and-butter market niches. But glance at the product lines of NetManage, WRQ, and fellow host-integration competitor AttachMate Corp., among others, and you’ll see an array of products that have evolved beyond the green screen terminal emulation tools that typically characterized the host integration market as recently as five years ago. Admittedly, this evolution may have been driven by necessity—for a stretch in the 1990’s, there wasn’t much, if any, licensing growth in the terminal emulation space—but these new products have also evolved in response to incipient demand from customers.

The upshot, says Havart-Simkin, is that leading-edge companies are tapping new Web services-friendly application architectures, such as J2EE and Microsoft Corp.’s .NET, to expose mainframe process as part of a service oriented-architecture (SOA). “Whereas five years ago, I could have said, ‘I’m not going to worry about the Web services architecture too much, because I can migrate this stuff off and move it to WebSphere or [BEA Systems Inc.’s] WebLogic,’ today the mentality is that these [host systems] are an asset to be revalued and taken forward as part of a services-oriented architecture,” he argues.

Although NetManage and other host integration vendors acknowledge that only a small percentage of customers are today building SOAs that involve their mainframe systems, they argue that as IT spending picks up, SOA investments will, too.

Winning New Workloads

Jeanette Stroud, vice president of product management for mainframe technology with software giant Computer Associates International Inc., says that most of the customers who were planning to phase out their mainframe systems have already done so. “The number of customers depending on the mainframe has gone down in the last several years, but the customers who have mainframes today have pretty well solidified,” she notes. “So if they’ve already made the investment and if they enjoy the qualities of service in terms of availability, performance, efficiencies, being able to support thousands of apps and end users all at the same time, most of them have recommitted to the mainframe in the last couple of years.”

Last year, market researcher Gartner Inc. found that although there was a lot of turn-over in the low-end—40 MIPS or less—mainframe space, most of the new mainframe systems sold by IBM were in the 80 to 1000 MIPS range. Customers with minimal investments in mainframe capacity tended to abandon the platform, even as customers with more significant investments added capacity. In fact, Gartner last year found that shops with more than 10,000 MIPS typically expanded their mainframe investments.

For her part, Stroud agrees with this assessment, and notes that even though fewer mainframes are being shipped these days, there’s more mainframe capacity to be had than ever before. “In many cases, the amount of capacity, and with the z990, particularly, the amount of memory or in-core storage that the customer can buy for his dollar, is much greater,” she points out.

At the same time, she says, IBM has successfully positioned the mainframe as a platform for new workloads, such as Java, WebSphere, and, of course, Linux. “IBM has done things to broaden the use of Java within z/OS, but now the other kind of new workload that has come to the mainframe platform, is of, course the Linux workload, and beyond server consolidation, we are seeing applications vendors and customers responding to this,” she says.

Gartner says that over the past two years, Linux-related uptake has helped to spur additional investments in mainframe capacity, accounting for 15 percent of the new MIPS shipped in 2001 and 20 percent in 2002 by its estimation.

In this regard, a mainframe Linux technologist with a global services vendor says Big Blue’s decision to standardize the price of its Integrated Facility for Linux (IFL) across all of its architectures has made it easier for customers to deploy Big Iron Linux. “Now, given that IBM has made IFLs the same [list] price across all their architectures, if we have existing boxes that would need to be upgraded, it won't matter what that architecture is. Anything completely new coming in would most likely be z990s, whether standard processors or IFLs,” he says.

Another pricing reform introduced by the Mainframe Charter was a reduction in the cost of mainframe memory, from around $520,000 per GB (according to RethinkIT analyst Barry Graham) to $10,000 per GB. This, too, was designed to spur uptake of Linux and other new workloads on mainframe systems, because such applications typically require more memory than more efficient mainframe workloads. “The 'new' applications that IBM is trying to attract to the mainframe are memory 'heavy', as they come from the Unix platform where 64-bit capability has existed for some time,” Graham wrote in a research bulletin last October. “Linux also 'eats up' memory at an incredible rate compared with traditional applications which have, by contrast, evolved over time from all environments where memory was always in short supply.”

Turning the Corner

As of early 2004, says IBM’s Mastrobattista, most of the capacity that Big Blue ships is earmarked for new workloads. “When we look at 1998, which was now about six years ago, at that point, only 15 percent of the MIPS that we shipped would have been in support of new workloads, so the remaining 85 percent would have been dealing with the more traditional online transaction processing, like CICS,” he notes. “That percentage is now well in excess of 70 percent, so in six years time, people are really looking now to the mainframe to be a more strategic part of what they’re now doing.”

IBM has committed to take the mainframe to even greater heights in the coming years. By the end of 2005, for example, Big Blue is expected to double the number of processor engines supported by the z990, from 32 to 64. That’s quite an accomplishment on any platform, but is all the more impressive in view of the fact that by last October, T-Rex was shipping with double the processor complement of IBM’s former flagship mainframe system, the z900.

Some users, as can be expected, are more enthusiastic than ever about the mainframe’s future. “I concur with IBM's opinion that the sky is the limit concerning mainframe growth,” says Bob Richards, a mainframe operator with a U.S.-based financial institution. Adds Richards: “The advent of the z990s, IBM's commitment with their ‘Mainframe Charter,’ continued growth in the adoption of Linux on S/390 architecture, and the fact that there appears to be no abatement in mainframe sales seem to confirm IBM's opinion and mine.”