HP-Oracle Itanium Suit: Divorce, Silicon Valley Style

If we had a page 6 gossip column, that's where this story would go. Its motivations and characters are all about the inside baseball of Silicon Valley's elite circles, about decisions made long ago about products that in some cases were brilliant and in others were just bad bets. Of course, there have been twists and turns that could happen only in Silicon Valley.

Let's start with where we are today. On June 4, Oracle and HP gave opening arguments in front of Santa Clara County Superior Court Judge James Kleinberg, who will decide the case in lieu of a jury. HP is claiming a breach of a contract negotiated between the two companies shortly after Mark Hurd was hired as co-president of Oracle in 2010. The contract's intent was twofold: First, it sought to limit some of the projects Hurd could work on; and second, it introduced a "corporate hug" that ensured that the two companies would continue to cooperate in the marketplace. It's the second part of the contract that HP says was breached. HP is seeking $4 billion for Oracle's decision not to port new versions of its database to run on Itanium-based hardware.

So that's where we are today, but how'd we get here?

The Disco Era

In the late 1980s, every computer company worth its salt had its own line of RISC processors. Sun had SPARC, IBM had POWER, DEC had Alpha, and HP had PA-RISC. Even Intel had the i860, which performed quite will in its time, and there were others too. These processors all came with the premise that Intel's x86 architecture (used in PCs) and Motorola's 6800 architecture (used in Macs and many Unix workstations) were outmoded and in serious need of replacement. Intel's and Motorola's chips had a complex instruction set, or CISC, design, which meant many instructions took more than one clock cycle to execute, in some cases many more. The new designs streamlined the instruction set and for more than a decade were the fastest chips available.

Companies such as Oracle had no choice but to support the range of microprocessors available in order to make its database ubiquitous. But like all markets, the fragmentation led to consolidation, and while that happened, Intel took advantage of Moore's Law, which saw gates per chip go from a few hundred thousand in the late '80s to millions in the '90s and now to billions. What seemed like a complex instruction set at one time was now child's play to implement. Not only is there room for lots of CPU cores on the chip, there's also room for memory management, FPUs, GPUs, and huge memory caches. Those and wide execution units make it possible to execute any instruction in a single clock cycle.

During the consolidation, HP was the biggest buyer of faltering hardware companies. It bought Apollo and Compaq, and Compaq had previously bought DEC and Tandem Computers. As a result, HP found itself with systems that ran Intel, Motorola, DEC/Alpha, as well as its own PA-RISC chips. That was far too many chips for the company's needs, so it started to end-of-life some of its product lines.

The Alpha and Motorola lines were the first to go, and HP was in a tough place with PA-RISC, too. The company thought it needed to make a bold move, and rather than continue down the RISC path, it went a different route. What if you could issue more instructions in a clock cycle and have them all execute in parallel? If you could, you'd get a speed advantage without having to resort to high clock rates, which imply high power usage. HP called it EPIC, or explicitly parallel instruction computing. The instructions were entered by packing many of them into a single "word" (64 bits in the case of Itanium, which permitted six instructions per word).

Global CIOs: A Site Just For You Visit InformationWeek's Global CIO -- our online community and information resource for CIOs operating in the global economy.

The theory sounded good, but HP didn't want to foot the bill for what was turning out to be a very expensive execution of the EPIC theory. The hardware was tough enough, but the real trick was making compilers smart enough to take full advantage of the EPIC architecture--that continues to be a problem today. To help with costs and expertise, HP started partnering with Intel on the technology as early as 1994, and it eventually handed the whole thing over to Intel with promises to buy Itanium chips well into the future.

Many predicted that the IA-64 architecture, as Intel dubbed it, would be the dominant chip architecture for high-end systems, but Moore's Law has a way of making fools of us all. Large caches and multicore designs did more for performance than the basic processor architecture could. So we are where we are today, with x86-64 chips performing on par with other available chips.

The Sonny & Cher Years

During that time, HP and Oracle were like Sonny & Cher, The Captain & Tennille, or any other duo you can think of that was inseparable--right up until the moment they separated. HP did the hardware, Oracle did the software, and they were a fine rival to IBM. The marriage hit its first major bump in the road when Oracle bought Sun Microsystems. Suddenly, HP hardware didn't look so attractive to the new Oracle/Sun.

The second bump came when the HP board, led by former Oracle president Ray Lane, ousted Mark Hurd over some expense reports and alleged illicit extracurricular activity. That ouster was quickly followed by the ultimate insult, HP's naming Leo Apotheker as Hurd's replacement. Apotheker had been at the helm of SAP, the German ERP software giant, during some unsavory clashes with Oracle, so his hiring was a glove to the face of Oracle CEO Larry Ellison.

Ellison didn't like the way HP had treated Hurd, his old friend from the Sonny & Cher days. So he hired Hurd to run sales and consulting, a move seen by HP as violating the non-compete clause in his HP contract. And so HP sued Oracle. The suit was seemingly resolved by a new contract between the companies that calls for, among other things, mutual support of products, which HP has taken to mean Oracle can't stop making its database for Itanium-based systems until HP itself declares them dead. And that's the case that Judge Kleinberg is now hearing. Oracle says it got wind that HP and Intel were going to pull the plug on Itanium anyway, so it was just acting in the interest of its customers and hoping that it might turn a few of them into Oracle/Sun customers.

You Hurt Me, I Hurt You Back

Customer interests aside, there's little doubt that the impetus for Oracle pulling support for Itanium was to make life hard for HP. HP's Integrity systems, which use Itanium, will be far less attractive to a great many customers if Oracle won't port new versions of its database to them. However, given the performance trajectory of Intel's x86 chips, it's reasonable to assume that Intel and HP were contemplating a way out, so while it happened in an ugly way, Oracle may have just accelerated a process that was going to play out over the next decade anyway.

In the trial's early going Ann Livermore, who negotiated the "breezy contract" between the two companies, along with Oracle's other co-president, Safra Katz, said that HP was completely blindsided by the press release which announced the discontinuation of Oracle support. She's also inferred that Katz seemed blindsided by it too, indicating that was cooked up by Ellison and possibly Hurd. Because the nature of the contract was relatively casual--intending to stress that HP and Oracle remained BFFs--Judge Kleinberg may look closely at the intent of the timing and way Oracle announced its discontinued support.

While Oracle is in better straits than HP, the two big winners from this skirmish have been and will continue to be IBM and Microsoft, which will be quite happy to work with customers that now have a bad taste for Oracle, HP, or both.

This is a case that would normally be settled by reasonable participants. But Oracle and HP aren't likely to be reasonable. HP is wounded and Oracle smells blood in the water, and so, oddly enough, each party sees this ongoing spectacle as being in its interests. Bottom line, I'd be planning my path away from Integrity systems, and if I could do it in a way that also got Oracle off my back, I'd think about doing that too.

Art Wittmann is director of InformationWeek Reports, a portfolio of decision-support tools and research reports. You can write to him at [email protected].