Mad Dog 21/21: Two Calves And Calves Not
December 1, 2014 Hesh Wiener
In Semitic mythology, Jews at the foot of Mount Sinai embraced an invisible God rather than one represented by a golden statue of a calf. From that pivotal moment, the followers of Moses and eventually the other people of the book, Christians and Muslims, vested their faith in software rather than hardware. A couple dozen centuries later, the resultant monotheistic civilization controls the bulk of the word’s wealth and power. This year, one of that civilization’s great industrial institutions, IBM, shifted the burden of its strategic trust from physical hardware to intangible software and cloud computing.
For many customers, and probably for most of them, IBM’s evolution is late rather than early. Users of IBM i technology have long since learned that Big Blue re-hosts their computing environment from time to time. In the past there has been some disruption (but not the same amount for every user) and a lot of progress (but not always as much improvement in price/performance as performance for legacy workloads). If IBM changed the i’s hardware platform again or even embarked on a course that promised changes every few years, committed users would adapt. They would embrace gains from whatever progress directly helped them. And some would grumble. Change might lead to defections, but for lively i shops, lack of change would present an even greater temptation to exit the IBM i universe. The System/34, System/36, and their families faded away when they were frozen, not while the equipment was evolving.
Mainframe customers are often more tightly bound to their hardware than i users because quite a few run ancient applications or use venerable patches that depend on the specifics of IBM’s instruction set and its implementation. Some of the oldest code run by mainframe shops is cloaked in mystery, its origins uncertain, its documentation little more than folklore, its behavior when run on a new platform capable of surprises. Nevertheless, mainframe shops have flocked to less costly IBM-compatible systems whenever they have had a chance, from the long-ago days of Amdahl and Hitachi PCM processors to the more recent FLEX and Hercules platforms. Amdahl and Hitachi built IBM-like hardware and wrote IBM-like firmware. FLEX and Hercules emulations run on X86 hardware. Users generally got all their code to work and only abandoned or avoided the non-IBM alternatives when Big Blue scared them off. So, if IBM introduced a new hardware generation and promised support for all or nearly all legacy code, as it has in the past, offering in return improvements in functionality, power and particularly value, users would stick with their big iron. As is the case with the IBM i, progress can keep the mainframe alive, while paralysis would kill it off.
The IBM Power platform that is used for AIX and Linux as well as the IBM i operating system changes a bit with each generation, more or less the way X86 and ARM processors do. Generally speaking, instruction sets get richer, functions that have become more important to users get additional improvements and characteristics that are connected to fundamental properties of chips, such as cell geometries, change as transistors and their interconnections become smaller.
In addition, for some applications, such as scientific computing, IBM has become much better at coupling its Power chips to coprocessors, coprocessors based on graphics processors from Nvidia. This change in the Power environment currently affects very large computers running enriched software, particularly supercomputers, but it is bound to seep into systems used by more ordinary folk, computers aimed at big but not gigantic jobs. Customers who can take advantage of the enriched hardware just love it; customers with more ordinary workloads like their association with computers aimed at the extremes of information processing and, most likely, hope some of the technologies pioneered by research labs and other exotic institutions will eventually give them an edge in meeting the challenges that arise in commercial and governmental settings.
So, with a customer base that is tied to the three server families that IBM has made over the years, how can the company walk away from chip fabrication? And how will customers react if IBM takes the next step away from hardware and decides to buy whole servers, not just their chips, from outside suppliers?
IBM thinks the outcome will be just fine for customers and for itself, as long as it can help customers get their work done the way they know how. Moreover, IBM has come to believe that it can do an even better job pleasing its customers and its shareholders if it can get everyone to believe in its abstracted offerings and let go of their attachment to the boxes in their glass houses and, if their workloads are hosted elsewhere, the physical facilities of their service bureaus.
This change is so dramatic that it really cannot be viewed as the embarkation of the good ship IBM on a new course. Rather, the focus of customers, shareholders, employees, suppliers, resellers, and every other interested party is on IBM’s leader, Virginia Rometty. She might be the most courageous, powerful and influential boss IBM has had since the days of Tom Watson, Jr., when IBM bet its survival on the System/360. She is already being compared to Lou Gerstner, who was put in charge during 1993 to save the company from itself, and did all that was expected of him and more.
In the eyes of customers, investors, and employees, Ginni Rometty already looms far larger than her immediate predecessor, Sam Palmisano. All the people and all the companies that depend on IBM are watching very closely to see which legacy products will continue to get emphasis and support and which ones will not. They are also wondering which executives will survive. All this uncertainty is bound to take its toll on customers, and thus on IBM itself.
When the biblical Jews were wandering in the desert, Moses, their leader, temporarily left them and climbed up Mount Sinai. While he was away he left his brother, Aaron, in charge. After 40 days, Moses returned, bringing with him laws, the Ten Commandments, which would form the foundation of a new culture. In his absence, he discovered to his dismay that many of the Jews had lost their faith in his invisible God. They had built a golden idol in the form of a calf and made it the focus of their ceremonies. Moses was furious, and pretty mad at his brother, Aaron, too.
In the Old Testament Aaron takes the blame and excuses his failure by explaining that the idol was wrong but that the alternative would have been discord between the factions that wanted the new hardware and those who were willing to stick with the software system Moses asserted was the true desire of the God they could not see. In the Koran, Aaron gets somewhat better treatment. A man names Samiri is said to have organized the construction of the golden calf and its placement at the center of religious activities.
In both accounts of the golden calf story, things get ugly. The Jews narrowed their focus to those who were members of the tribe of Levi. Moses and Aaron wore Levi genes. The Levis had stuck with Moses and his faith and when rallied by their leader turned on the idolaters and, as so often happens in the Old Testament, smote the others. The Islamist version of the story also ends up with a lot of smiting, too. Faith in a single God is a hallmark of Judaism; so orders one of the Ten Commandments. Muslims, too, say Allah is the one and only; they call diversion from that belief the sin of Shirk, and it can have severe consequences.
If you happen to think IBM and its Cravathian tribe of lawyers was tough on plug-compatible mainframe manufacturers and others who copied or coopted IBM’s intellectual property, well, you can see there’s some important precedent for Big Blue’s harsh and deadly response.
Soon, I expect Rometty and her executive team will turn from the negative, such as relinquishing the manufacture of hardware, and point their faithful customers toward a future reached via networks. The contents of the glass house may be replaced by communications apparatus. The corporate data center, whether based on mainframes, Unix servers, or IBM i machines, will fade away. In its place will be Rometty’s vision of the computing future: the promised LAN.