The IBM Systems Agenda: iB(M)

by Timothy Prickett Morgan

Be careful what you ask for, because the universe has a sense of humor. It often gives you something that sounds like what you asked for, but when you examine very carefully what you got, it is not exactly what you wanted. It's a little bit off to one side or another, twisted in an ironic way, or just plain wrong because the universe doesn't listen to us--we are supposed to listen to it. What applies to universes applies to IT suppliers: Be careful what you ask for.

It is with this in mind that I have been mulling over a little-discussed, non-announcement by IBM that was called "Project Polaris" internally and is known as the IBM Systems Agenda to the outside world. I have mentioned this in passing since the COMMON user group meeting last month, as have a bunch of IBMers. Frank Soltis, the iSeries chief architect, even joked at COMMON that the audience was not "in the know" about the announcement, which is essentially a five-year plan for Big Blue's four server lines. The reason why they didn't know anything about it was simple: IBM has not, in the sense that it normally does these things, made an announcement. It has kinda mumbled a few things here and there, and some time when no one was looking, probably after the "Danu" System z9 mainframes were launched and maybe just around COMMON, it launched a Web site that talks a lot, and sometimes with some details, about exactly what Project Polaris is all about. (By the way, more than a couple of people who play Code Name Bingo like I do got the name of the new z9 mainframe--code-named "Danu" after the Irish earth goddess and mother of the pantheon of Irish gods--with Project Polaris.) The z9 is, at least in IBM's marketing presentations, an embodiment of the goals and ideals of Polaris. But Polaris is not the code-name of that mainframe, but is rather the North Star, as in the star to steer by. As in the iSeries.

Like most central governments, IBM likes to think in five-year plans. (Why five is a magical number is unclear, but things tend to fall in units of five and 10 naturally. Whether this is an effect of our base 10 numbering system or the cause of us having developed a base 10 numbering system in the first place is the subject of a different story.) Most of us remember the last five-year plan. That was "Project Mach 1," with that number relating to the speed it takes to break the sound barrier, which brought us the eServer over-arching brand name for IBM's servers and the renaming of the System/390, AS/400, RS/6000, Netfinity, and NUMA-Q systems to the zSeries, iSeries, pSeries, and xSeries servers. (Yes, the NUMA-Qs did not live as a product line, but a lot of the engineering that the former Sequent Computer Systems created to make its NUMA architecture for Intel-based servers lives on in the "Summit" and "Hurricane" generations of high-end xSeries servers and in the midrange Power5-based i5 and p5 servers, which use similar technology.)

There was a lot of weeping and gnashing of teeth about the whole eServer thing, if you will remember, and that is probably why when IBM launched the System z9 on July 26--the day we launched our new Big Iron newsletter for mainframe customers--it did not make a lot of noise about the IBM Systems Agenda. IBM's experience with eServer was not exactly a positive one, so you can't blame it. As silly as some aspects of the eServer rebranding were, some of it made good sense.

With Mach 1, IBM made a bunch of commitments, the first of which was to more aggressively start sharing technologies among its server units. It even formed a more tightly coupled Server Group (now known as the Systems and Technology Group, and almost certainly to soon be known as the Systems Group), and had a single head of development for the whole server organization. The first two people to hold those roles came from IBM's mainframe labs--Ross Mauri initially and, today, Vijay Lund. That isn't precedent; it's just that IBM's mainframes generally have most of its most advanced hardware technology in them. The Power4-based "Regatta" servers absolutely buried the mainframe in terms of performance, with about a 3-to-1 ratio for online transaction processing workloads on the largest single system image, but in terms of availability, reliability, batch performance, CMOS packaging technology, and on-demand utility computing features, the IBM mainframe is still ahead of any Power-based machine. At least for a while longer. Power6 may change that, and if not, Power7 certainly will.

The sharing between IBM units has deepened over time because, to put it bluntly, IBM has four different and largely incompatible server lines and cannot fund their development as if they were four competitors. When mainframes cost $100,000 and a MIPS and the PC server was a joke, IBM could get away with this. But since the 1992 recession, when the mainframe business imploded, IBM has had to do some belt tightening and some soul searching. The AS/400 and RS/6000 lines were already on a path of convergence thanks to the 1991 PowerPC alliance. But that effort got kicked into high gear in 1995 when Motorola totally messed up the 64-bit PowerPC 630 design and IBM Rochester had to save the day with its own 64-bit "Muskie" PowerPC AS chip--which is the great granddaddy of the Power5 we have today. Not coincidentally, in 1995 IBM considered rebranding its servers, which was not coincidentally five years before Project Mach 1. By 1997, IBM started talking about "e-business" and the high-end AS/400 and RS/6000 servers--the "Raven" S70 and the "Blackbird" S7A--which were essentially the same machines with slightly different I/O and memory subsystems. At that point, eServer was happening, it just didn't have a name.

But the Mach 1 project was about more than that. The idea behind the eServers was to give all of IBM's four server brands common characteristics: support for Linux, support for logical partitions, support for capacity on demand computing, and the highest reliability and availability Big Blue could pump into the platform.

The IBM Systems Agenda seems to be eServer on steroids, without all the fanfare and hoopla. (IBM could have called it the Hidden Agenda, in fact, since it really didn't promote the idea at all.) IBM quietly named the Danu mainframe the System z9--no slash like mainframes of old, but systems nonetheless--and last week, with the launch of the Power5+ versions of the p5 and OpenPower machines, the pSeries machines were called the System p5. And, yes, it seems likely that the next big announcements in the spring or early summer for any Power5+-based iSeries machines will see a naming shift to something called the System i5. I have no idea what IBM plans to do with its xSeries, since System x86-EM64T sounds silly, and it sells Opterons, too. Maybe System x64 will last until a few decades from now, when we move to 256-bit processors? And what about the BladeCenter blade servers, which have no system name associated with them. System BC? (What would that mean, an ancient computer, like an abacus?) It could be worse. If IBM had chosen the name of Hewlett-Packard's blades, which are called BladeSystem, for its blades, then Big Blue would have to call them System BS, which would probably be one of the truer statements in IT these days. But I digress. Often, and with enthusiasm.

The mantra coming out of Big Blue with the IBM Systems Agenda will have a familiar ring to it, but there are some new elements that IBM is adding in. The so-called systems IBM is talking about will have three common attributes: virtualize everything, a commitment to openness, and collaborate to innovate. (Those are IBM's terms, not mine. I would have just said, Open Up, Collaborate, Virtualize, or something like that.)

The virtualization story is just what happens when you mix the logical partitioning and capacity on demand aspects of the eServer approach and bake it for a few years. The Virtualization Engine software at the heart of the iSeries and pSeries today is based heavily on the technologies that the Rochester labs created to bring mainframe-like partitioning to the AS/400 many years ago. IBM's AIX Unix had to be gutted to ride on top of this virtualization hypervisor and hardware functions had to be added to its Power processors and systems to better support dynamic logical partitioning. IBM has played around with doing partitioning on X86 servers, but has donated its efforts to the Xen open source project and has strongly backed market leader VMware. And mainframes, no matter what you call them, have their own way of doing things. But IBM is trying to make them all easier to manage, and over the long haul will equip its IBM Director software with the capabilities to manage virtual environments and the provisioning and workload management feature of specific operating systems across all of its server platforms in the most consistent way possible. Storage virtualization will also play into this strategy.

As far as openness, this means more than just selling open systems, as IBM has done since 1990, or helping develop and sell Linux, as IBM has done since 2000. IBM has inadvertently created some great open standards--the PC AT is the best example, and one that helped spawn Compaq and Dell--as well as purposefully created others, such as the SQL programming language standard or the Eclipse integrated development environment. IBM gives a lot of lip service to open standards, and talks a lot these days about its Blades.org blade server community and its Power.org Power architecture, and it clearly supports a lot of open standards in its systems. But let's be real here. All IT vendors use open standards as weapons, just like they used proprietary standards in the past. Show me an open standard, and I will show you one vendor trying to muscle a lot of others.

The collaborate to innovate theme of the IBM Systems Agenda is a reminder to IBM that it has to share its technologies across its server brands, to give them all as similar technology and capability as possible. This makes perfect sense, and it has been going on for a decade in earnest and really since 1991 when the PowerPC architecture was rammed down the throats of the AS/400 and RS/6000 units. At the time, the AS/400 business had just turned in its best year ever (and a better year it has never had since then), and the RS/6000 business was but a wee bairn.

More than a few executives have talked about the server industry moving over the "sweet spot" of the iSeries in recent months, how an integrated, easy to use, business system is the kind of thing that customers want to buy. I couldn't agree more. But I want to remind IBM that the "i" in iSeries was for "innovation" and "integration," and that was supposed to be the machine's unique value proposition.

Here comes the "be careful what you ask for" moment.

What the IBM Systems Agenda sounds like to a lot of ears in the OS/400 community is making IBM's other machines more like the iSeries, and I think if you proposed it that way, they would not exactly be happy about it. A lot of people think that IBM should be selling the iSeries harder than it is, not making other machines more like the iSeries. About half of IBM's customer base is comprised of iSeries customers, after all. The idea should be to double that iSeries business, not give all systems--z9, p5, and x64--the attributes of the i5. (That is what I mean by iB(M) in the title: little "i" for iSeries, capital 'B" since IBM is all about Business, and "M" in parentheses because the specific machines are not really of a concern.)

Then again, the zSeries zealots have probably been saying the same thing about the iSeries and pSeries people for the past two five-year plans, and the iSeries and pSeries people have probably been saying the same thing about the xSeries team. Sharing is difficult, but this strategy proves one thing: IBM has come to the conclusion that customers cannot easily get rid of their legacy mainframe, OS/400, and AIX workloads and that it can make more money by keeping these customers happy and letting its machines interoperate in virtualized environments. This strategy is a lot easier than trying to converge down to one IBM platform. That is for sure--at least until Power6 or Power7 or Power8, if you believe the rumors running around.

RELATED STORY

international Business (machines)