Mad Dog 21/21: Noshing Like Cronus
February 3, 2014 Hesh Wiener
IBM is a modern day Titan. In Greek mythology, the Titans were children of Uranus who preceded the Olympian Gods. If an observer were to pick one Titan to compare to IBM, it would be Cronus, whom the Romans called Saturn. Cronus ate five of his children in an effort to preserve his power, the way IBM has eaten the compatible mainframes and Power and mainframe emulators its technology spawned. A sixth child of Cronus, Zeus, eluded that fate, as the PC did in IBM’s case. In the end, Zeus deposed Cronus.
IBM has done a fantastic job of preserving the uniqueness of its mainframe and Power systems. But it hasn’t always been easy. IBM’s victory over mainframe clone makers and emulation system vendors involved an arsenal of technical and legal strategies. Even so, one viable mainframe emulator is still freely available: the Hercules system. Hercules can run 31-bit and 64-bit mainframe apps including entire IBM and mainframe Linux operating systems, middleware, compilers and actual applications programs. Users cannot, however, get licenses for production grade IBM software; Big Blue insists on licensing its code only to its own hardware or the handful of alternatives that are grandfathered in, such as Flex-ES. There are basically two groups of Hercules users these days: Experimenters who are happy to run ancient IBM software that predates Big Blue’s decision to take its code private and underground users who have figured out how to put some or all their work on systems using unlicensed code.
To understand just how dramatically IBM would be impacted if it didn’t eat its mainframe-imitating children, all one has to do is check out the Raspberry Pi-based Hercules mainframe that a hobbyist can build for a hundred bucks, more or less. It will boot ancient MVS. It will boot mainframe Linux. It will boot stuff that IBM says you cannot legally run, such as current VM and VSE. Raspberry Pi is an ARM computer that runs Linux. Hercules can live inside a suitable Linux, such as the Debian distro. IBM’s software can execute inside that Hercules.
If the Raspberry Pi, based on processor technology that’s now getting old, can boot a mainframe OS, it’s a safe bet a processor based on newer, faster ARM system-on-chip designs, of which there are plenty, would have enough power to do real commercial work . . . if IBM permitted it.
Hercules on X86 hardware is vastly more powerful and can yield systems with hundreds of MIPS, more power than many legacy shops have on their floor or in the mainframe partitions they rent from mainframe service bureaus. It is pretty clear that IBM’s restrictions that allow it to preserve its mainframe monopoly are absolutely necessary if it wants to stay in the mainframe processor business.
Critics may argue that IBM should have long since let go of its grip on hardware and put all its revenue eggs in the software and services baskets, but IBM’s management obviously disagrees. However, now that the X86 business has followed IBM’s disk drives, printers, and PCs to the exit, IBM’s remaining hardware manufacturing operations, basically the fabrication of mainframes and Power servers, may become difficult to preserve. This is the case even though a decade ago IBM found a way to eliminate the most serious technological threat to its Power technology, but did it a bit late. Before IBM eliminated it, that threat cost Big Blue its most important chip customer, Apple.
IBM did no better than Cronus, but no worse.
Cronus ate his kids right after they were born. He had been told by his father and mother, Uranus and Gaia, that his sons would depose him, which turned out to be the case. It was a harsh solution, but that was the way things were done in the Titans’ realm. And eating his children was only one of the things Cronus did to grab and maintain his power.
Cronus became the top Titan by reducing his father, Uranus, from a powerful, active force to an important but impotent spirit of the heavens. He did this by castrating him with a scythe. Later, Cronus was defeated by his son Zeus, who, according to some myths, got Cronus to regurgitate Zeus’ siblings or, according to other myths, cut Cronus open to free his siblings. In addition, Zeus rescued some other relatives that Cronus had imprisoned, earning their support.
As you can see, Zeus was quite a lively character who became the alpha deity of Olympus. As for Cronus, after losing his preeminence, he remained an important figure in what was back then the services business, agriculture, where he was depicted wielding a scythe, the harvesting implement he used to cut off his father’s testicles.
Never wasteful, the mythologists add that after Cronus cut off Uranus’s testicles, he tossed them into the sea. Where they hit the water the surface boiled up a pillow of sea foam, out of which appeared, fully formed in stunning womanhood, Aphrodite, memorably depicted in Botticelli’s Birth of Venus.
As astonishing as this story of succession may be, a similarly stunning tale emerges from the history of Apple, which twice changed processor architecture, porting or supporting legacy software, while gaining rather than losing its customer base.
Apple’s first platform change occurred in 1994, when Apple introduced the first Power Mac as its successor to prior models that used Motorola 68000 series CPU chips. Apple systems software and applications for the Power models at first included emulation code that allowed new machines to support legacy programs. In 1996, Apple ceased building 68K-based machines and gradually wound down support for the old hardware as it introduced successively more capable Power products.
Apple quickly became the most visible user or Power chips outside IBM and the only company with a successful line of client machines based on IBM’s power architecture. As part of a complex licensing and production arrangement, Apple bought Power chips based on IBM intellectual property from Motorola, which fabbed the glass and encapsulated it for later assembly by Apple and its manufacturing subcontractors.
Meanwhile, IBM continued making X86 PCs until late 2004, when it sold its PC division to Lenovo. IBM only briefly added Power-based PCs to its product line in the early 1990s–just before almost immediately killing them–and never put its muscle behind the sale of Apple clients to its enterprise customers.
IBM’s lack of skin in the game may have contributed to Apple’s decision, twelve years after moving to Power, to move once again, this time to Intel X86 technology on a hardware platform that was capable of running not only Apple’s Mac OS but also running Microsoft Windows. (Remember, Windows used to run on the Power platform, and that was the plan for the PowerPC Alliance from the beginning. Microsoft did not stick to the deal just like IBM didn’t.) Windows on a Mac could fly solo on a machine set up with a choice of systems made at startup or in a window that supported many apps. (This Windows capability was popular among tech support personnel and a relative handful of users but not at all of interest to the vast majority of Mac users, who preferred true Mac apps and the user evolving interface style that gave the Mac its distinctive and attractive personality.)
When in 2006, two years after IBM left the PC business and a dozen years after adopting Power chips, Apple switched to Intel-based processors from Power circuitry, it faced even more daunting technical and marketing challenges than it had in the mid-1990s. Computer hardware and software had become vastly more complex. The Internet had reshaped client computing around interactivity that included intense interactions with media libraries, dynamic publications, and peer-to-peer communications services. It was clear to Apple that Intel had not only pulled ahead of Power when it came to computing capabilities and price/performance, but that its chip designs were going to leave Power in the dust. Apple felt that if it stayed with Power it would be crushed. Still, Apple had to come up with a migration plan that preserved the trust of its users. Enter Transitive.
Transitive was a software company that found particularly efficient and accurate ways to translate Power code to Intel code on the fly. Transitive called this specific capability Rosetta. The emulation tax on legacy Apple apps running on the new Intel boxes was affordable at first. It would quickly become less of a disadvantage as Intel learned how to make faster chips, software creators learned how to rewrite and retune their code for the new platform and users became more adept at getting the most out of the Intel-based Apple systems.
It has taken the better part of a decade for the impact of IBM’s failure to develop, in a timely fashion, Power processors with the pricing and efficiency Apple needed, for the IBM customer base and the investing public to appreciate how decrepit IBM’s hardware business has become. Some are puzzled by IBM’s decision to sell its X86 business rather than go the other way by porting its mainframe and Power systems to X86.
Such a move would once have been relatively easy for IBM, because in 2008 Big Blue acquired Transitive, the company from which Apple acquired its Power-on-X86 emulation technology. IBM basically bought it to destroy it (but not its intellectual property rights) and to prevent others from hurting Big Blue the way Apple did when it switched processor chip suppliers. IBM may have lost Transitive’s skills and talent, as it has not unveiled a Sparc-to-Power translation. But it is impossible for an outsider, and maybe even most insiders, to guess. There might be other factors shaping IBM’s decision to go after the former Sun Microsystems base using traditional sales tactics rather than leading-edge technical tricks. Or it might be that Sun users who are particularly unhappy with their systems and looking for change would be more inclined to head for the competitive X86 market than to go from the Sun frying pan to the IBM fire.
Another mystery, perhaps a more important one, surrounds the current status of Transitive technology owned by Apple. For those with a high opinion of Apple’s institutional intelligence, it would not be surprising to find out that Apple acquired lasting and living rights to Transitive’s offerings. Such rights would allow Apple to deploy its translation weapon again, perhaps to enable to move software from Intel to ARM, or, as Intel’s mobile chips improve, from ARM to Intel. If Apple doesn’t have such rights, it must have great regrets.
As for Intel, it constantly worries about pretty much everything. That’s its corporate nature. It nevertheless hopes to hold focus and understand the value of sticking to its priorities. Right now Intel’s biggest worry is competition for X86 from ARM chips. Power isn’t an issue. If Intel can vanquish ARM in client markets and keep its advantage in servers, it will stay far ahead of Power when it comes to delivering hardware value. If Intel loses to ARM or is forced to cede a significant amount of territory, Power users will see that they are not one but two major steps behind.
If recent trends continue, at some point Power and mainframe users may respond to IBM’s entreaties in a fashion quite different from the way Leda responded to that swan, the one that was really Zeus in disguise. They might tell IBM’s persistent sales reps to go pluck a duck.