HP, Intel, and Oracle Gang Up on IBM Mainframes
Published: October 26, 2006
by Timothy Prickett Morgan
As part of the festivities at the Oracle OpenWorld customer event in San Francisco this week, Hewlett-Packard announced that it was teaming up with chip maker Intel and Oracle, which has the most popular relational database on Unix servers and which also has middleware and many different application software suites, to try to persuade IBM mainframe shops that it is now time to move to new architectures and new applications.
This particular offering, called the Application Modernization Initiative, is being spearheaded by HP Services, not its Enterprise Storage and Server group, which is responsible for making and selling the company's ProLiant and Integrity server lines and related storage products. While HP is relying heavily on its Integrity line of Itanium-based servers--as well as its HP-UX Unix variant, that operating system's Virtual Server Environment logical partitioning and MC ServiceGuard clustering, and the Oracle middleware and database stack and its clustering capabilities--to make a case with mainframe shops that they are spending too much money on their mainframes, the difference between this offering and past mainframe replacement deals is that this one leads with services. That means lots of assessment and hand-holding, which mainframe shops always appreciate.
According to Paul Evans, worldwide director of application modernization services within HP Services, Oracle, Intel, and HP executives got together a few months before the "Montecito" dual-core Itanium 9000 series processors were announced in late July, and they all shared some insights on their respective mainframe alternative programs. "What we agreed on was that mainframe customers were saying that they are interested in alternatives because of dwindling skill sets and the desire to implement services oriented architectures, but at the same time they were also worried about the risks associated with making a change."
That latter bit is a big deal, even after you consider the substantial premium that mainframe shops pay for hardware, software, and services. Evans says that even though mainframe customers acknowledge that they are spending 70 percent of their IT budgets on maintenance of legacy applications and the underlying platforms that support them, mainframe shops large and small are adverse to risks. (After all, if they were not, they would have moved to Unix boxes a decade ago.) HP reckons that about 10 percent of the mainframe shops are already on the move, that 10 percent will never move, and that the remaining 80 percent are interested in moving, but want to better quantify the risks associated with moving.
"What we need to do is show mainframe shops that if they move, they are not going to be doing final test and assembly in their own data centers," explains Evans. HP had already set up five SOA Excellence centers around the globe, and in fact has already launched various application modernization services back in November 2005 through HP Services. Now, Oracle and Intel are helping to build a legacy application modernization practice in these centers, and one of the first steps is the obligatory reference architecture. Such reference architectures are the blue prints that show how to build up a mainframe-class data center on Unix iron.
"The biggest thing we need to work through is perception," explains John Pickett, who runs the mainframe alternative program at HP Services. "Very few customers have a comprehensive view of their applications. And our goal is to find the applications that have high value but low technical quality that are running on mainframes and help customers move these." Old, monolithic COBOL applications that access flat-file IMS, VSAM, or IDMS databases are a prime example. "A lot of these applications have functionality built into them that companies really do not need to code any more--such as handling printing and forms, to give one example--that you buy a package to do these days."
Pickett says that HP's target mainframe customers for its Application Modernization Initiative have 1,500 MIPS or more of aggregate mainframe processing capacity on site. HP believes that 80 percent of the mainframe sites in the world have less than 1,500 MIPS of processing capacity, and that there are between 15,000 to 16,000 mainframe footprints in the world. HP says further that only about 3 percent of the customer base has more than 4,000 MIPS on site. But don't get the wrong idea. HP Services is interested in smaller mainframe shops, and will even provide similar services to customers using IBM's AS/400 and iSeries servers, too. (In fact, Pickett says HP is working on a deal to port 2,100 AS/400 and iSeries servers at one customer to other platforms, and this is a customer who moved from mainframes many years ago to save money.)
HP Services has done a bunch of these mainframe migrations in the past decade, and some of them have been quite big. HP has just finished a migration of a bank in Southeast Asia that involved 7,500 MIPS of mainframes that drove 800 branches and 40,000 bank agents. The COBOL applications were ported using tools from Clarity and Tmax Soft; these tools can emulate the COBOL-CICS environment, which rode atop a three-node Oracle 10g RAC cluster running on Integrity iron. The bank says that it will save $20 million over four years and will get a return on its investment for the port in 18 months.
To get those 80 percent of mainframe shops that are risk adverse to look at their options, HP Services is offering a mainframe site assessment and rationalization plan for $50,000, which is pocket change for big mainframe shops. (Mainframes cost millions of dollars, and big mainframe users can spend millions of dollars a year on core z/OS, IMS, DB2, CICS, and related software.) "We want to get in quickly and help explain what is possible--and that this is not as death-defying as they might think," says Evans.
IBM, of course, is not taking this sitting down, and will almost certainly start wheeling and dealing when the HP Services salespeople get their feet in the data center doors. Two weeks ago, IBM announced that it would offer System z mainframe shops a $22,500 to $250,000 rebate when they acquired a new System z9 mainframe and Oracle 10g databases for the mainframe. And also at Oracle OpenWorld this week, IBM said that it was working to get the Oracle Applications ERP suite ported to Linux on the mainframe. Moving to Linux on a mainframe to support Oracle applications will cut the mainframe hardware and software bill significantly, since Linux is a lot cheaper than z/OS and IBM charges about a quarter for mainframe engines running Linux as it does for those running z/OS.
But at that point, is it really a mainframe or not? And once you are ready to move to Linux, Integrity servers and other Itanium-based machines offer much better bang for the buck than even the reduced-price Linux engines on System z9 mainframes. If IBM wants to keep its native software and hardware alive against competitive platforms, it has to either cut prices and compete or believe that its customers are too risk adverse to try to move their applications, and that they will use SOA technologies to try to modernize without messing with their current systems too much. The latter strategy is a safer bet for IBM profits in the short term than the former, so it is not hard to guess what IBM will do.
Bang for the Buck: Big Iron Boxes, Even Bigger Bucks
A Joint Assault on the Mainframe Hardware Market
Intel Aims Dual-Core Itaniums at RISC, Mainframe Servers
HP Gears Up for Montecito Itanium Shipments