Oracle Goes Back to IBM’s Roots with Sun Deal Done
February 1, 2010 Timothy Prickett Morgan
I know it is hard to believe, but Oracle, which is now a systems provider thanks to the $7.4 billion acquisition of the former Sun Microsystems, actually wasn’t kidding when it said it wanted to be in the systems business. Forget that Oracle was really only initially interested, according to rumors, in Java and Solaris. Somewhere, the top brass at Oracle came to the conclusion that maybe getting into hardware might be the right move after all. And emulating IBM–the Big Blue that dominated the early years of commercial computing, not the services-obsessed behemoth you know today–was the way to future and dependable profits.
Last week, at the Oracle-Sun coming out party–and I mean Sun, not Sun Microsystems, because Larry Ellison, Oracle’s co-founder and chief executive officer, doesn’t do anything “Micro” and the Microsystems has been dropped from the brand name that is now slapped on the division that sells x64 and Sparc servers as well as storage, switches, and Solaris–the top brass at Oracle talked for six hours about the glorious future that lay ahead. It is a future that many of you, in the System/3X and AS/400 midrange base, and maybe those of you with experience with the System/360 and System/370 mainframes that predate them, will recognize in your rear view mirror.
The deal closed on Tuesday, and Charles Phillips, one of the two co-presidents at Oracle, set the parameters during the Oracle-Sun product rollout for the systems that Oracle will be building in the coming years as it seeks to extend what it offers and its control over the IT industry. “It’s been about building complete systems for Oracle for a very, very long time,” Phillips explained. “But the definition is changing.”
Silicon Valley–and the Hudson Valley too, for that matter–has gotten stinking rich by creating best-of breed-solutions in the many tiers of the IT stack–servers, storage, networking, operating systems, middleware, and applications–and having them grind away against each other in the market. But integrating these parts is tough, and you end up with brittle systems that are difficult to change. This makes IT systems “very unpredictable, very unreliable,” Phillips said. “That is why it is so expensive to maintain these systems.” He added that some things just can’t be done after a system is up and running, that you can only do them at the front end, and that changes in systems are the main reason for support calls at Oracle.
But that is not how it has always been. Because of the complexity, cost, and incompatibility of hardware platforms, the first mainframe and midrange machines were completely or mostly integrated. In fact, it took several antitrust lawsuits against IBM to bust the platforms open enough to make a place for Oracle and myriad other IT suppliers in the early computer business. Oracle relishes control perhaps more than anything else, because in control there is profits. And that is the reason why Phillips referred to that IBM as “the gold standard of the computing industry.” And it is why Oracle wants to create completely integrated machines, from chip to disk to applications, that have predictive, pro-active, and constant maintenance.
“To do that, we’re going to spend a lot of money,” Phillips said, saying that Oracle would boost its research and development spending by nearly 54 percent to $4.3 billion, which means a large part of the remaining Sun R&D budget (after some pretty steep cuts in 2009) will be intact. There is, of course, the $7.4 billion Oracle paid to get Sun in the first place, too, but it wasn’t that expensive because Sun had a bunch of cash on hand. It is hard to say how much, since Sun didn’t put out financials for the quarter ended in December 2009, and never will. At the end of September 2009, Sun had $1.76 billion in cash and was burning about $120 million a quarter; it was shelling out on the order of $1.6 billion in R&D, about half of what it did only a few years ago. So basically, for about $5.8 billion and the maintenance of the Sun R&D budget, which the struggling server maker could actually afford, Oracle got ahold of a basically breakeven company raking in maybe $11 billion to $12 billion in revenues. Oh, and one that Oracle thinks will bring in $1.5 billion in operating profits in the next 12 months and $2 billion in the year after that because it has Oracle standing behind it, engineering systems, and not slashing the employee count to the bone.
Of course, to be fair, former Sun president and chief executive officer, Jonathan Schwartz, cut 10,500 employees from the time he started talking to IBM about a deal in November 2008 to when the Oracle deal was finalized last week, cutting the employee base to somewhere around 22,500 if I have the initial numbers right. So Ellison gets to look like the hero, saying that Sun will only cut a one or two thousand jobs as part of the merger (he gave different answers to different people), and will be hiring 2,000 hardware and software engineers as well as sales and marketing people.
“We’re excited to be in all of these new businesses with these fantastic people from Sun,” Ellison said at the event, saying that integration will offer huge value to customers. And in a dig at IBM, he used IBM’s own history against it. “It’s not that this hasn’t been done before,” he said referring to the IBM run by Thomas Watson, Jr. “That strategy made IBM the most important company in the history of Earth, so we kind of like that strategy. So we are going to adopt it.”
If Larry, the Sun King, wasn’t interested in hardware back in early 2009 when he was trying to do a deal with other hardware suppliers to take over Sun and ditch its hardware biz, he sure is now. And the way Oracle is going to do that is simple: it is going to sell Sparc SMP systems made by Fujitsu and clusters of Sparc and X64 machines made by Oracle and tune its entire software stack to run like gangbusters on those boxes. It is going to simplify Sun’s supply chain and cut down on its product line–no more chasing the volume X64 server business in SMB accounts–its distribution centers, and its manufacturing operations. It is taking over the top 4,000 Sun accounts and it will sell into those directly, just like IBM does with the top 5,000 accounts that give Big Blue so much of its profits and account control. And it is completely backtracking on the old Sun’s failed strategy to become a volume player in the X64 server racket.
“We’re not too interested in being in the commodity X86 Windows market,” Phillips said dismissively. “Other companies, Dell and so forth, are good at that. Let them do it.”
Well, they aren’t as good as they used to be at making money on X64 boxes. (There are no X86 boxes, since 32-bit chips haven’t been sold for years, and X86-64 is too long, which is why many of us in the industry say X64.)
Before joining Oracle, and no doubt at Oracle’s urging, Sun killed off its 16-core UltraSparc-RK “Rock” processors and their “Supernova” systems, which were supposed to be here last fall; well, actually the fall before that and then they were delayed. Permanently, as it turns out. It looks like Rock was dead sometime before June 2009, but Sun never confirmed anything about it except to take a write-off for chip development. So going forward, Oracle will be relying on a the existing Sparc T series multicore roadmap, which has four generations of chips in the pipeline, including the “Rainbow Falls” 16-core, 128-thread processors due around the middle of the year, and three follow-ons.
In standalone systems with one to four chips, these Sparc T3 chips are perfectly fine if your applications like threads and don’t need a lot of cache memory or clocks (the will only run at 1.67 GHz, apparently). In clusters, they might be just fine on workloads like Oracle databases and applications. The Fujitsu Sparc Enterprise M machines have quad-core chips running as high as 2.88 GHz, and will get a goose in the next year or so to 3 GHz (whoopee!), which is not exactly exciting. IBM’s eight-core Power7 chips are probably going to cream them on any test, and entry machines could give the Sparc T boxes a run for the money, too, with 32 threads per socket and at least twice the clock speeds as the Rainbow Falls chips. They damned well better.
But Oracle being Oracle, it is really interested in clusters of lots of machines and now that it owns chip designers and its own operating system, pushing all the guts that make clusters support online transaction processing and data warehousing workloads down closer to the iron. This has been a 15-year journey for Oracle, and it is a journey that IBM has not been willing or able to take, except for Parallel Sysplex clustering for DB2 workloads on mainframes and some dabbling with DB2 Multisystem on the AS/400 for the past 15 years to not much effect. (Something I have complained about frequently.) Yes, I know about DB2 PureScale OLTP clustering for AIX, which was speed-launched last fall after Oracle and Sun put out the Exadata V2 Database Machine. DB2 PureScale was supposed to be ready in December, and nary a peep from IBM.
If IBM isn’t peeping, Ellison is growing like a chanticleer about the advantages Oracle has over IBM when it comes to OLTP workloads as well as data warehousing workloads. Here’s what Ellison had to say when someone at the event suggest IBM had the same assets and could come right back at Oracle:
“They do not have the same assets. The Oracle database scales out; IBM DB2 for Unix does not. Let me see. How many servers can IBM put together for an OLTP application on what used to be called DB2 UDB–IBM has two databases, one on mainframes, one on Unix, totally different products–let’s see, how many of those can they group together?
“You would have thought, years ago, that IBM would have come up with its own Database Machine. I mean, it’s
I don’t hate to say I told you so. I agree with Ellison, and I told you so, IBM. If Ellison is right, then the industry is going to swing wildly away from where it is now, with services-laden, hand-cobbled, mixed systems from myriad hardware and software parts suppliers
The problem is bigger than clustered databases. It is that Oracle is the first company since the 1980s to have an integrated hardware, systems software, and application software stack. Yeah, the stack is a bit messy right now. So was IBM in 1993.
Maybe we can get Oracle to call its line of integrated machines them the Business Technology/500s, the follow-ons to the heritage of the Application System/400s? Oracle seems to be listening.