In-Depth

State of the Mainframe, Part II: The Capacity Conundrum

Some mainframe pros say that next-generation workloads (such as zLinux and J2EE) point the way to false destinies

• By Stephen Swoyer
• 02/07/2006

The mainframe has endured, to be sure, but just how it will continue to endure is the subject of some contention—and not a little bit of consternation.

Many Big Iron visionaries say the mainstreaming of next-generation workloads such as zLinux (nee, Linux/390) and J2EE will be the key to the mainframe’s long-term viability. Some mainframe traditionalists take quite the opposite view, however—namely, that zLinux and J2EE point the way to false destinies.

Most mainframe watchers, of course, are somewhere in the middle.

Mark Post, deputy project manager for the SHARE Linux project, is a pretty big next-gen workload booster. This doesn’t mean Post turns a blind eye to the value of traditional COBOL applications—just that (instead) he believes that Linux/390 is here to stay, and, if anything, will become more scalable and functional over time. “For the shops that just want to keep on running COBOL applications forever, I don't have a whole lot of sympathy. Anyone who hasn't gotten the message that they can't keep doing things the way they've always done them probably isn't long for this world, unless they have some sort of monopoly position in their market,” Post says.

Why make the move to Linux/390 and other next-gen workloads? It’s all about the dollars, Post argues. “[T]here are at least a couple of COBOL compilers for Linux/390. [Customers] could port their applications, run them on an IFL, and reduce their z/OS capacity charges substantially,” he says.

It’s not as if IBM Corp. gives away its Integrated Facility for Linux (IFL) engines for zSeries, but it’s certainly true that Big Blue’s Linux IFLs are priced to move. In this respect, Post says, customers can run Linux workloads on mainframe hardware for a fraction of the cost of blue-blooded z/OS COBOL applications.

Capacity Comparisons

It’s precisely on this point that mainframe veteran Ted Tiefeld, a principal with Big Iron consultancy Tiefeld & Associates, disagrees. “If IBM was not distorting mainframe usage by its pricing schemes, anyone who studied the reality of business applications would rightly conclude that Linux and/or Unix is the worst thing to do to a mainframe,” he argues.

Tiefeld says he isn’t a knee-jerk reactionary, either: he’s trained with an IBM business partner on Big Iron Linux and has explored other mainframe operating environments over the last few years. “The result of that training was that I ended up exploring more about the mainframe, and became more interested in zVM and TPF, with some lesser interest in zOS. This is because I discovered that the processing model of Unix/Linux is a poor match for the strengths of a mainframe in terms of workload characteristics of business processing, and serves only to guarantee that IBM will sell more engine MIPS than ever before.”

The upshot, says Tiefeld, is that the still immature Linux operating system can’t husband system capacity and resources nearly as efficiently as z/OS, TPF, VSE, and other traditional mainframe operating environments. “IBM loves Linux because it is such an inefficient utilizer of mainframe strengths,” he comments. “It guarantees more hardware revenue. If IBM would price traditional mainframe workloads like Linux-based workloads, no one in their right mind would migrate from traditional workloads, save for [the] faddish followerism that afflicts the computer industry and computer education establishment.”

IBM officials say it’s ludicrous to expect Linux—even zLinux—to have feature and function parity with z/OS. Hence the reason for the premium IBM charges for most native z/OS workloads. “We do believe the [traditional] mainframe provides significant added value over the competition, and so we want to make sure that it is priced appropriately,” says Colette Martin, zSeries program director for IBM. “As customers move from z900 to z990 to z9, and as they move from z/OS 1.4 to 1.5 to 1.7 … those customers who stay the most current [on mainframe hardware and operating environments] have seen very dramatic price/performance improvements in z/OS workloads of all kinds.”

At the same time, Martin and other IBMers point out that zLinux users can take advantage of many features that are unique to the zSeries mainframe, including HiperSockets connectivity to CICS, DB2, and IMS running on z/OS, as well as mainframe quality-of-service and reliability features.

A Pragmatist’s Take

Bill Miller, vice-president of BMC Software Corp.’s information management (IM) business unit and a long-time mainframe market watcher, describes himself as a mainframe realist. Yes, he acknowledges, zLinux and other new workloads still aren’t as mature as z/OS or as efficient as VSE, TPF, and other bread-and-butter Big Iron environments. Even so, Miller suggests, customers seem to be diving into next-generation workloads with the expectation that things will get better.

“The number one pain point we are hearing [from customers] is SQL performance, because what they’re telling us is happening is that more and more applications are moving to the mainframe, but they’re being coded in the distributed world before they move to the mainframe,” he comments. “And in our world, with mainframe applications and SQL performance tools, you can very easily and effectively take advantage of those tools to improve your coding. What customers are interested in is taking advantage of that with these new workloads coming on board because the coding, and what they do with the applications, is maybe not as sturdy or as reliable.”

Ditto for mainframe J2EE, Miller says—largely because of IBM’s über-affordable zSeries Application Assist Processor (zAAP) for Java. “WebSphere is becoming very hot right now. IBM is pushing it hard, so we’re actually investing quite a bit in our WebSphere product,” Miller explains. “Customers are saying, ‘We’ve got to have WebSphere tomorrow, we’ve got to have [WebSphere] MQ tomorrow, we need it right away on the mainframe part.’

“That kind of reinforces what I said about SQL performance. We’re seeing the traditional applications are still there on the mainframe, and they’re going up a little bit, and more frequently, we’re seeing these new workloads, which are the result of Linux and WebSphere adoption.”

Manifest Destiny

In the long term, some mainframe market watchers say, the net effect of Big Blue’s pricing policies will be to steer more and more customers toward next-gen workloads—or (conversely) to force traditional customers to upgrade to new versions of z/OS at something just short of a glacial pace.

When a Big Iron shop purchases Linux IFL(s), for example, it doesn’t have to re-purchase an IFL each time it upgrades to Big Blue’s latest and greatest mainframe hardware (e.g., from z900 to z990 to z9). Customers do have to pay Linux licensing, maintenance, or support costs (to Red Hat, Novell, or other vendors; or to business partners and services providers, for example), but such costs amount (right now, anyway) to a fraction of similar z/OS costs. The net effect of such a policy is to provide customers with ever more bang for their buck as mainframe systems become ever more powerful.

This value proposition isn’t quite as straightforward for z/OS shops. A z/OS 1.3 customer that wants to deploy z/OS 1.4 will pay much more, on average, than a zLinux shop would to upgrade to the latest version of Novell’s SuSE Linux Enterprise Server. For some organizations, this amounts to a double whammy: not only must they foot the bill—$250,000 or more—for new mainframe hardware, but they must also pay a not-insignificant price for new z/OS software.

“A lot of clients are hanging as far back as they can to avoid having to upgrade and pay the price [new z/OS license charges], while not losing the ability to easily migrate to the ‘64-bit-only’ version [of z/OS],” says Post, whose 9-to-5 job is with a prominent outsourcing services company. “IBM delayed that version several times, due to loud complaints from people that weren't prepared to pay for new hardware and software simultaneously.”

---

In an upcoming column we’ll examine mainframe training: Who should pay—and why? Big Blue isn’t the only one with an interest in the future of the mainframe platform—but why are American companies so loathe to foot the bill for mainframe (and most other technology) training?