Distributed Computing, Specialty Engines Bring Companies Back to the Mainframe
The cost of distributed complexity—which is also measured in air conditioned BTUs and kilowatts per hour—is about to get even more expensive
- By Stephen Swoyer
- July 11, 2006
The success of a new presidential administration is often measured by the performance during its first 100 days in office. Not surprisingly, then, most presidents try to pack an awful lot into those first days in office.
The first 100 days of IBM Corp.’s new System z GM Jim Stallings were extraordinarily eventful. Stallings stepped into the role in January and promptly announced an ambitious plan to visit 100 mainframe customers in 100 days. Dismiss it as a publicity stunt if you like, but Stallings says all of that jetsetting was a valuable experience—and not just as a means to rack up frequent flyer miles.
“It was amazing to me, it was astounding. What was really clear was that all of the numbers were saying that the mainframe was back … but when I visited these 100 customers, they were telling me about even further growth, and these weren’t just big [companies], I have intentionally met with new customers that are buying a mainframe for the first time,” he says.
Stallings says he visited new mainframe customers in the U.S. as well as Belgium, Australia, and, of course, China and India. The mainframe has always offered compelling value for small- and medium-sized customers, Stallings says, but its value proposition has at times been overshadowed by its daunting upfront cost. That isn’t necessarily an issue anymore for these customers, he argues.
“In April, I went to China and announced our new business-class mainframe. If I’m [a mid-market] customer, it is at a price range that I can talk about. This puts this value at a point where I can acquire it,” Stallings says.
The timing couldn’t be better, he argues. In fact, Stallings links the rise of offshore outsourcing with the growth in complexity—or with a coming to terms with the cost of the growth in complexity—of the highly distributed enterprise. “Companies are beginning to realize that the cost of maintaining [their distributed infrastructure] is getting to be huge. It’s driven a lot of companies to consider, ‘I can maintain this app and develop it much cheaper offshore,‘ and it’s driven a lot of that work offshore,” he comments.
“When you compare that to the mainframe, we’ve got a whole different dynamic. We now have a very competitive story [to tell] and it’s driving workloads off of the Unix and Intel platforms and back on to the mainframe. We set this up so that we positioned ourselves not only to be attractive at initial acquisition, but also over time [for ongoing TCO].”
The mainframe’s new entry-level price point isn’t the only arrow in its suddenly brimming quiver, Stallings maintains. Its raft of specialty engines—which include a bevy of workload-specific engines, such as the Integrated Facility for Linux (IFL), the zSeries Application Assist Processor (zAAP), and, most recently, the zSeries Integrated Information Processor (zIIP)—are helping drive uptake of new workloads on zSeries, too.
“We announced the zIIP engine in February, we [made it available in May], so we’re still in the early days of it, but we’re already seeing tremendous interest. If you look at our success with the other [workload engines], the IFL has outperformed our expectations. We have at least 3,500 IFLs out in the market right now, and we’ve got some customers who’ve installed literally hundreds of IFLs on the mainframe, on a mainframe,” Stallings comments.
By the same token, Big Blue’s Java-ready zAAP engine, which has been available for almost two years now, appears to have occasioned a surge in WebSphere-on-zOS deployments, if anecdotal accounts are correct.
With zIIP, IBM’s strategy was to encourage organizations to bring it all back home—at least with respect to RDBMS processing—to the mainframe. With only two months of general availability under zIIP’s belt, it’s too early to tell if that plan is succeeding, Stallings stresses, but he’s optimistic that zIIP, like its predecessors, will help drive workload deployments on zSeries.
“You’ve seen the surveys that show that if you take a billion-dollar company, most of them have multiple databases. In some cases, they have over 100 different data structures, so what some of them are considering doing is consolidating. They go through a process that says, what’s the best platform to consolidate on? When IDC surveyed CIOs, 30 percent of the time they said they were going to migrate their database management systems to z/OS,” he comments. “The zIIP makes it extremely easy for a customer that has a mainframe to not get a different database management system [provided they’re running DB2]—they can just move it to [DB2 on] z/OS. They’ll get the same performance, if not better.”
Stallings says Big Blue plans to introduce additional workload engines, too --starting perhaps with a new engine optimized for security processing. “There’s a secure data hub that a lot of customers are thinking about. Because of all of the security concerns, especially with banking and health care, they’re beginning to think about ‘How do I create this security management infrastructure for my environment, and is there a security management platform [designed] to manage passcodes and keys and all of the security globally?’” he explains.
Stallings cites a few other trends he believes are working in the mainframe’s favor. Consolidation, after all, is a product of the cost of complexity, and the cost of complexity reached a tipping point—i.e., the impetus to consolidate—only when other factors, such as the cost of data center real estate and the cost of providing power for the operation and cooling of the data center, became prohibitively expensive. That’s Big Iron’s ace in the hole, Stallings argues.
“The more you load on the mainframe, because of virtualization, and because it can run at utilization rates that are 85 to 90 percent, you get this trifecta on the mainframe. There’s study after study that proves it is the most powerful platform when it comes to cost, efficiency, and performance,” he indicates.
“If you take a workload on a Unix box, or an Intel box, when an application is idle, the utilization rates are driven down into the low teens and sometimes single digits. You’re still paying for the application, you’re still paying for the air conditioning, you’re still paying for the power to run all of that. With the mainframe, what you gain back in terms of real estate and utilization is astounding, because utilization is so high and virtualization will enable you to run the thing at multiple workloads at 90 percent or more all the time.”