In-Depth
System z: What IBM Still Needs to Do
IBM has done a lot to help revitalize the mainframe, Big Iron technologists say, but there’s still much work to be done
Over the last 36 months, IBM Corp. has done a lot to help recast the mainframe as a much more affordable proposition—for new and existing customers alike. Big Blue delivered three new specialty engines—the Integrated Facility for Linux (IFL), zSeries Application Assist Processor (zAAP), and zSeries Integrated Information Processor (zIIP)—and increased its efforts to bring new mainframe technologists into the fold, announced a significant investment in mainframe simplification—especially in the application dev and systems management arenas—and also shipped its least expensive mainframe systems to date.
For these and other moves, most mainframe pros give credit where credit is due.
"The new z9 systems announcement, with a significantly lower entry price was a big shot in the arm for the mainframe server market. This, coupled with IBM's renewed interest and commitment to, z/VM is making the mainframe a very attractive application platform again. The future looks brighter, I think," observes Dave Jones, a principal with VM software specialist V/Soft Software Inc.
Not that Jones is completely sanguine about the mainframe status quo. In a sense, he says, IBM has done what it had to do to help counterbalance the cost-prohibitive licensing practices of third-party ISVs. That being said, he believes there’s still work to be done.
"What we would like to see is the removal from the marketplace for the need for such specialty engines. They exist only because third-party z/OS ISVs kept increasing their license fees, based on installed MIPS or MSU capacity, even if that capacity was being used by other operating systems like z/VM," Jones comments. "IBM was forced into creating such specialty engines like the IFL merely to make adding new, non-z/OS-based workloads economically attractive to mainframe shops, since they do not count towards a system's MSU rating as determined by third-party ISVs."
Nor is the specialty engine approach a silver bullet, Jones points out. "[Specialty processors] add another bit of complexity to the management of zSeries processors—for example, an LPAR cannot have a mix of IFLs and standard engines in its definitions."
Other mainframe pros concur. "These specialty engines so far are primarily a software license tool for z/OS installations and may be an attractive solution for those problems," says Rob van der Heij, a mainframe technologist with VM performance management specialist Velocity Software Inc. "In general, splitting up your processing capacity into units with specific affinity makes it harder to do capacity planning and tuning. There's a good and a bad side of that, but I don't expect IBM to do these things for z/VM as well, so it does not bother me yet."
Emulation is the Thing
Elsewhere, mainframe technologists say, and in spite of the emergence of sub-$100,000 mainframe systems, Big Blue still seems disinclined to really grapple with the prohibitive costs of premium Big Iron hardware. "Our biggest problem as a small—though publicly-traded and by no means tiny—ISV is that of an affordable development platform. IBM seems to consider its ISV partners as a profit center, and at the moment we have no modern system to develop on at a price we can come close to affording," says a Big Iron technologist with a mainframe-centric ISV who asked that his name not be used.
This mainframe booster, like many of his colleagues, thinks the last, best hope for affordable mainframe silicon lies in scalable emulation technology.
Given Big Blue’s historical ambivalence toward such approaches and its long-standing refusal to sanction high-end, 64-bit mainframe emulation, he doubts whether any viable solution will ever be brought to market. After all, IBM last year allowed an existing licensing agreement with emulation specialist Fundamental Software Inc. to lapse and (nearly) simultaneously targeted another emulation vendor (Platform Solutions) with an IP lawsuit (see http://www.esj.com/enterprise/article.aspx?EditorialsID=2327).
"There are all kinds of rumors about a coming Power-based emulated solution, but so far no action and no announcement. This, I feel, makes the barriers to entry for small ISVs extreme, particularly for those who might otherwise consider porting apps to z/OS or z/VM," he concludes. "When an ISV—or prospective [ISV]—can buy a high-end UNIX or Windows development platform for a couple [of thousand dollars], there is precious little incentive to develop for [zArchitecture]."
Ray Mullins, a mainframe technologist with a mainframe ISV that he asked not be named, picks up on this idea.
"I'm … concerned about the lapsing of the agreement with Fundamental Software and the FLEX-ES platform licensing. Although it's rumored that IBM will have a small system for its PWD developers in the future, those same rumors say that it is not near," he indicates. "IBM is making it damn near impossible for a small, independent developer without much cash to work on creating products for operating systems that don't have Linux in its name."
So does Stephen Frazier, a mainframe pro with a state government agency based in the Southwest. "IBM's refusal to allow 64-bit systems to run on commercial FLEX-ES machines is one of the worst and most short-sighted decisions in IBM's history," he contends. "This means that small businesses will never be able to use mainframes. As small businesses grow into large businesses, they will not be IBM mainframe customers. IBM's moves in 2006 (if not reversed) will lead to the destruction of the company.
For these and other reasons, many Big Iron pros think Big Blue can and needs to do more to make the mainframe more affordable—and increase its potential marketability. "Overall, I like the z9 introductions in 2006, but I'd like to see IBM move towards getting back SMBs who are being forced to migrate to inferior boxes and operating systems … due to costs—this goes for both IBM and ISVs," Mullins concludes. "I'm also glad to see z/VSE get some updating—I think IBM needs to put more back in z/VSE, though, to keep the SMBs."