State of the Mainframe (First in a Three-Part Series)
In spite of their differences, mainframers seem to agree on a few important points, although even these may surprise you.
Mainframe technologists are an opinionated bunch—with good reason. Most have weathered precipitous ups and downs in the daunting mainframe market waters of the last two decades. Mainframe pros, perhaps more than any other platform technologists, have seen their fair share of Next Big Things. With few exceptions, most become Next Big Busts. The mainframe, for its part, endures.
Not surprisingly, mainframe pros have a pragmatic take on the Big-Iron rebirth of the last few years. Nearly all mainframe technologists we spoke with are encouraged by Big Iron’s sudden reversal of fortune, but some also cite infrastructure issues that may be all but intractable.
The important point is that not all mainframers agree about just what ails their favorite platform. Some—such as Ben Tiefeld, a principal with consultancy Tiefeld and Associates—believe that zLinux and other next-generation workloads are false gods, so to speak. Others, including Mark Post, deputy project manager for the SHARE Linux Project, think that Big Iron pros who cling to z/OS, VSE, and other blue blood operating environments are only forestalling the inevitable.
In spite of their differences, most mainframers seem to agree on a few important points—namely, that mainframe hardware and (particularly) mainframe software are still far too expensive for most adopters, and that mainframe training could be a lot more affordable and accessible.
Capacity At What Cost?
“A mainframe is like having an aircraft carrier in your back yard. If you don't have one, you certainly don't need one. But if you do have one, you should probably do something with it, because it's not going anywhere,” says Joe Poole, a mainframe systems programmer with a prominent retailer.
This is what Poole and other mainframe enthusiasts call Big Iron Staying Power. But Staying Power of this kind comes at a price, Poole cautions. Mainframe software costs, especially for traditional products (COBOL, Assembly, and etc.), remain a prohibitively expensive proposition.
Even Poole’s employer—a true blue Big Iron shop—is having difficulty footing the mainframe software bill. Poole’s CIO is exploring the idea of migrating the company’s existing COBOL applications to Linux running on zSeries hardware. Poole stresses that his company remains committed to the mainframe for the long haul, however. It attempted to purchase a new z9 system from IBM last quarter, for example, but wasn’t willing to wait several months for Big Blue to supply it with one. In the end, Poole & Co. had to “settle” for a z990 instead.
“A mainframe is a tough sell unless you have a current z/OS workload and have some servers that can be brought up on z/VM and Linux. Without that workload, you'll have a hard time justifying a six-digit monthly software invoice,” he notes.
Software costs aren’t the most immediate bulwark to broader mainframe adoption, however. Big Blue has done a lot to make some mainframe workloads—such as zLinux, J2EE, and even z/VM—more affordable. But there doesn’t seem to be much that IBM can do about the exorbitant cost of mainframe hardware—at least relative to commodity Intel servers, along with increasingly inexpensive Unix systems, too.
You Paid How Much for Your New Baby Z?
Not everyone takes this view, however. Some Big Iron pros say Big Blue doesn’t have a competitive mainframe offering in the low-end—100 MIPS and under—segment. While sub-100 MIPS mainframe shipments probably won’t ever achieve commodity volumes, some Big Iron boosters think there may nevertheless be an overlooked opportunity there.
“I do think that the entry-level mainframe systems are far too expensive. IBM has pretty much ceded the below-100 MIPS market to the FLEX/ES suppliers. I think that's a mistake, but I don't know that anyone is going to convince IBM otherwise,” says SHARE’s Post. In his 9-to-5 existence, Post works for a prominent global services firm. He says his employer flatly rules out using FLEX/ES hardware—even if a customer doesn’t qualify for Big Blue’s smallest mainframe systems. “I know my company won't even consider using FLEX/ES, even if we have clients that want a dedicated machine, but don't come anywhere near needing an entry-level z800.”
At some point, Big Blue’s mainframe market strategy seems to have changed, Post argues. He’s not sure what to make of this shift. “IBM used to have the long-term view that small clients eventually turn into bigger clients. They seem to have lost them somewhere along the line.”
Not Everyone’s a (Knee-Jerk) Critic
Few Big Iron technologists are knee-jerk critics, of course. In fact, some—such as Post—can point to more than a few positive things IBM has done to help the mainframe keep on keeping on. At the same time, however, Post and other pros we spoke with say Big Blue’s work isn’t yet done.
“[T]here's the software costs, which [are] also pretty excessive at the low end. I think IBM needs to have some sort of ‘ramped’ price structure that allows an organization to get in fairly cheaply, with the bulk of the costs coming later in the contract. I also think they need to simply lower the price per MIP at the low end. This will help them capture some of the small clients who want to become bigger clients,” he argues.
Post also likes what IBM has done to make z/VM more affordable. “IBM may be inflexible on their z/OS (and perhaps VSE) pricing, but I know for a fact that the pricing for z/VM has been cut in half two or three times in the last four years, and converted to a one-time-charge per processor. That's nearly unheard of for IBM,” he points out. “In speaking with various IBM'ers … they would like to get to the point where everyone that orders a mainframe would get z/VM included, just as PR/SM [LPAR] is included today.”
This wouldn’t necessarily be unprecedented, Post concedes, because PR/SM is based on VM from some time ago. Regardless, he says, it can’t hurt IBM to more aggressively fluff its z/VM feathers: “They rightly see z/VM as the best in class virtualization technology and think that IBM as a whole should be promoting it more. I agree. People in the Intel world are just starting to realize what VMWare can buy them, and have no idea that z/VM has a 30-year head start, with all the additional functionality and maturity that brings.”
Poole, for his part, is strongly in favor of a single, sweeping change: More competitive pricing for z/OS workloads. While pricing parity for z/OS apps and zLinux apps may be out of the question—IBM says it has a right to charge a premium for the reliability, availability, security, and performance of z/OS on zSeries—near-parity pricing might be slightly more realistic.
Instead, Poole says, IBM typically increases its prices for CICS, IMS, DB2, and other bread-and-butter workloads from year to year.
“We go through [the same thing] every year. How can we reduce the software budget to compensate for the price increases for the next version of CICS or IMS or DB2? COBOL is about the only thing that makes sense on z/OS, whether it's in a CICS region or running batch,” he says. “Java doesn't belong there, with an assist engine or not. Mainframes really are reasonable as a hardware commodity when you consider everything they are capable of—until you install the software. IBM's software division is the one causing the most problems when it comes to budgets, and making a big profit to boot.”
In the next several weeks Enterprise Strategies will be venturing into more turbulent waters—namely, whether those who continue to power their enterprises solely (or mostly) with native z/OS, VSE, and other stalwarts of the mainframe establishment are charting a direct course for obsolescence.
We’ll also 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?