It's been almost 15 years since Microsoft jilted IBM's OS/2 at the altar, kicking Big Blue while it was already down (financially), and exacerbating the old school company's global state of disarray -- one that outsider (IBM's first ever) Lou Gerstner managed to turn around after he was brought on board in 1993. One year later, a group of Gerstner's handpicked technical advisors identified Sun's Java as the strategic technology on which some of IBM's fortunes could be protected or regained from Bill Gates, who, with a roadmap for Windows NT then firmly under his belt, was beginning to turn his attention towards one of IBM's most coveted domains: its data centers.
In the decade since, IBM has downplayed any insinuation that it's been seeking revenge for the way Microsoft defeated it and OS/2, often highlighting its continued sales of Windows desktops, notebooks, and servers as evidence of a healthy partnership between the two companies. But insiders know better. Call it what you want -- good engineering, great marketing, or foul play -- Windows dominated desktops and notebooks and was vanquishing its competitors on the server front (mainly Novell, but the Unixes too), which meant that IBM, and other systems manufacturers, were left with no choice but to bow to Sir Gates. For many, it came as no surprise that, once Linux showed promise as a way to derail Windows' momentum on the server front, IBM would throw its weight behind the upstart OS as well as other industry currents (eg: open source) that stood a chance of undermining Microsoft.
Over these years, as a student of the on-again off-again relationship between IBM and Microsoft, I have written many times that the relationship's status, invariably, can be connected to one's desire to take away the other's customers, or, at the very least, disadvantage the competitor in some way. When the relationship is off, it's almost as if one, the other, or both have thrown down some gauntlet. For example, IBM's recent backing of an EU crackdown on Microsoft makes the company's current Microsoft disposition pretty clear. When the relationship is on, the rivals have a mutual goal. For example, when Microsoft and IBM got together to form the Web Services Interoperability organization (the WS-I), lack of interoperability between IBM's Java-based Web services solutions and Microsoft's .NET-based Web services solutions was not only a fundamental impediment to Web services adoption, but also a barrier to easily substituting one company's solutions for the other's. It's a relationship that I've described before as being Eastwood-Van Cleefish in nature.
But, in addition to IBM's recent backing of the EU, is IBM making other moves at Microsoft's jugular? Taken together, do these moves signal a more concerted effort on IBM's part to further distance itself from Microsoft, an effort that could ultimately avenge the death of OS/2? Let's take into consideration some of the more recent news coming out of IBM.
Lenovo takes over IBM's desktops and notebooks: Although IBM will maintain an investment in the company that it sold the majority of its desktop/notebook interests to, IBM's divestiture had to have wiped out the majority of Windows licenses (across desktops, servers, and notebooks) that the company was selling. This move alone has resulted in a significant reduction of IBM's dependence on Microsoft.
IBM's big 64-Bit Windows Chill: This is supposedly a momentous week for Microsoft. At the company's hardware confab (WinHEC), Microsoft is officially celebrating the release of the Server and XP versions of Windows that support the AMD64 architecture found in AMD and Intel's hybrid 32/64 bit microprocessors. Meanwhile, a 64-bit version of Windows that supports Intel's 64-bit pure-breed IA-64 architecture (found in Itanium-based systems) has been out for a while. While IBM sells servers that support the three modes of Windows (straight 32-bit, hybrid 32/64-bit, and pure 64 bit), a recent news story by News.com's Stephen Shankland (see: Sources: IBM ditching Itanium altogether) has put IBM's long-term commitment to Itanium in question. Reports Shankland in the story's opening paragraph:
IBM is not only discontinuing support for Intel's Itanium processor in a new generation of server technology, it's going a step further, dropping Itanium servers from the product line altogether, sources familiar with the situation said.
The story, however, was unconfirmed by IBM. So, I contacted the server folks at IBM for clarification and they were indeed very coy about their disposition regarding Itanium. According to IBM spokesperson Tim Willeford, the company "has no announcements regarding future Itanium-based products." The context of the conversation was my theory, based on Shankland's report, that an IBM pull-back on Itanium also equates to a pull-back on the pure 64-bit version of Windows Server.
Incorporated into that theory is the fact that if customers want that sort of mid-sized iron, IBM will be happy to provide it to them in the way of Linux running on its 64-bit Power5 processor-based OpenPower Servers. As reported in a January 2004 News.com story, IBM's Linux on Power was simply an initiative in 2003 but was converted to a group with revenue responsibility in 2004. In that same story, Brian Connors, IBM's vice president of Linux on Power, said that IBM hopes to make Power servers available at the same cost as those using Intel processors. Fast forward to 2005, and not only is IBM's OpenPower initiative well underway, but Linux distributors like Red Hat are lining up behind it.
True to IBM's traditional downplaying (but not denial) of any tie-cutting with Microsoft, Willeford responded with a reminder that sales of Windows Server still make up the majority of IBM's Xeon-based server sales. Thanks for the party-line, Tim, but let's be honest: Xeon-based Windows servers make up too lucrative a segment for any server maker to turn its back on, regardless of how badly a company might want to break free of the ties that bind it.
My take: If any part of the world is heading in the direction of pure 64-bit computing -- particularly datacenters -- don't expect IBM to lend Microsoft a helping hand there anytime soon. This is a definite shift.
And if you were worried about Linux ports of your favorite server apps on those OpenPower servers: IBM has you covered there as well. Given the lack of native ports of your favorite server apps to any 64-bit version of Linux, normally, heading in some sort of "we're cold turkey on 64-bit Windows direction" would leave a company like IBM in little-iron no man's land. What good are OpenPower servers with barely a handful of applications that can take advantage of them? Enter IBM's Chiphopper, announced at this year's LinuxWorld. Earlier this year, Chiphopper was explained to me in-depth by IBM's worldwide Linux chief Scott Handy. (For you're viewing pleasure, I did a Whiteboard session on Chiphopper too.)
In a nutshell, Chiphopper is an IBM technology that takes care of all of the heavy lifting when server software companies like Oracle and SAP decide they want to port their x86/Linux-based software to not just Linux running on Power5-based systems, but also Linux running on IBM's mainframes. Referring back to Red Hat's support of Power-based systems, here's one dandy quote that says it all: "We’ve already chiphopped, if you will, Red Hat and SuSE." Putting the OpenPower servers in place was simultaneously both an anti-IA64 and anti-Microsoft move. Greasing the wheels of application and OS portability to all of IBM's iron (including OpenPower servers) is the icing on the cake. IBM may dispute the connection as part of some master plan to take back from Microsoft what Microsoft took from it in the 1990s (aka: avenging OS/2's death), but I'll bet you that somewhere on some whiteboard at IBM, the connection was being made and that the execs who were nodding in approval were tasting Microsoft's blood.
Just in case you're not convinced: By itself, right- or downsizing to low-cost Linux boxes has proven to be a very compelling approach for many businesses. But it really gets interesting when you can consolidate those servers for the sake of big wins like easier server management, power savings, on-the-fly provisioning, and maximization of server resource utilization.
Today, VMWare rises to the top as a technology that supports the virtualization of both Windows and Linux in a way that one box can take on the appearance of multiple physical servers. Recently, VMWare (a subsidiary of EMC) even announced that it's expanding into the 64-bit realm. In Vanderpool, Intel has its own virtualization technologies on tap. And for Windows, Microsoft has Virtual Server. But why wait as these solutions come to market from vendors that have a limited amount of experience in virtualizing big iron machines when you can get it from grandaddy of virtualization--IBM?
Earlier his year, News.com's Stephen Shankland reported that IBM had very quietly open sourced the code behind its Research Hypervisor, otherwise known as rHype. (Given such a low-key approach and IBM's downplaying of tie-cutting with Microsoft, one has to ask if IBM is trying to fly below the radar.) In a virtualized environment, a hypervisor is the layer of technology that sits between a server's hardware and the multiple operating systems running on it, basically supervising the virtualization process and making sure each of the virtual servers can peacefully co-exist with each other. Officially, Big Blue says it has no plans to productize rHype. On the other hand, as a puzzle piece to IBM's overall ambitions, it should be noted that one of rHype's distinguishing features is that it works with IBM's Power processors, x86 processors, as well as the Cell microprocessor being codeveloped by IBM, Sony and Toshiba.
Perhaps the most interesting bit in Shankland's report was this: "Because rHype uses the same interfaces as [IBM's] commercial hypervisor, Linux doesn't have to be modified to run on an rHype-Power foundation. With rHype on x86 chips, Linux must be modified to work."
Admittedly, in the grander scheme of things, rHype doesn't come across as the sort of gauntlet that some of IBM's other moves may appear to be. But, taken together, Chiphopper and rHype add up to an interesting scenario where Linux-based OpenPower servers can be fully virtualized so that multiple, chiphopped instances of Linux and the chiphopped applications they support can run side-by-side on a low-cost, high performance 64-bit server that's been carved up by a mainframe-class virtualization technology from the company that practically invented the idea. At the very least, on paper, this scenario sounds pretty good when you try to mock up something comparable in the Windows-Itanium world.
Are more IBM puzzle pieces than ever now in place for Big Blue to conduct a major offensive against Redmond? Will the company finally avenge the death of OS/2? It sure seems like things are heading in that direction. Use our Talkback below to let me know what you think.