In-Depth

Behind Teradata’s Split with NCR

NCR and Teradata officials left a good deal unresolved last week, especially the question of who gets what.

• By Stephen Swoyer
• 01/17/2007

Will NCR Corp.’s divestment of its data warehousing division produce a Teradata that is leaner, meaner, and keener than ever to compete? NCR stockholders certainly think so. So, too, do many industry watchers. Indeed: given Teradata’s historical leadership in the high-end data warehousing segment, its success as an independent entity would seem to have the stamping of a fait accompli.

Nevertheless, NCR and Teradata officials left a good deal unresolved last week—starting first and foremost with who gets what. In the NCR universe, Teradata isn’t exactly an island unto itself: in the vast majority of cases, in fact, NCR-branded servers power Teradata’s best-of-breed Database and Warehouse. More to the point, Teradata’s special sauce—i.e., its Massively Parallel Processing (MPP) capability—is enabled by means of NCR’s proprietary BYNET interconnect. So it’s a no-brainer then: NCR gets the server business, right?

Unfortunately, the situation isn’t as clear-cut as that, according to sources familiar with NCR sales and marketing. Because while NCR does provide Teradata’s hardware special sauce, Teradata in turn accounts for most of its parent company’s midrange and high-end server sales.

“I believe that NCR does not actually sell midrange file servers and UNIX boxes anymore. As I understand it, NCR’s file server business is an ‘exited’ business, but ongoing support for existing product stays within NCR,” says a source familiar with NCR/Teradata who spoke on background.

Just what does NCR bring to the table, hardware-wise? Plenty, say data warehousing professionals who are familiar with Teradata.

“I can't see Teradata being viable without the MPP hardware to go with it. If they became a software-only company, their database would only have appeal to existing customers. After all, without the hardware, it becomes just another database system with no clear differentiator in the market,” says author Nick Galemmo, a data warehousing consultant, and veteran data warehouse architect, with consultancy Headstrong Corp.

True, Galemmo concedes, Teradata Database is optimized for analytic queries, but Teradata’s scalable, continuously available, shared-nothing MPP underpinnings are what give it most of its juice. “If you remove the MPP hardware, Teradata is equivalent to DB2 or Oracle. They all have means to optimize the environment for analytic queries, and in my opinion, of the three, Oracle does the best job with star schema,” he argues.

Author and Teradata veteran Jill Dyche, a partner with Baseline Consulting, agrees. To be sure, she says, Teradata can run on platforms other than NCR servers and NCR’s flavor of Unix—Windows 2000, Windows Server 2003, and SuSE Enterprise Linux are all supported—but very few customers actually choose to do so.

“Teradata can indeed run on non-NCR hardware … [but] you can really only leverage the powerful MPP capabilities via the Teradata-specific configuration,” she indicates. Dyche says she’s familiar with at least one customer who hosts Teradata Warehouse on non-NCR hardware; the catch, of course, is that this customer is also contractually prohibited from discussing his experience. In the vast majority of cases, Dyche argues, the MPP special sauce is what first draws customers to Teradata. “For a truly ‘shared-nothing’ environment, you’ve gotta go with a Teradata box, which is what the vast majority of Teradata customers do,” she concludes.

In some respects, says Mark Madsen, a data warehousing consultant and a member of TDWI’s Extended Research Collaborative, the hardware question might be a moot issue. “This commodity hardware and architecture leapfrogging is what always nips at the heels of specialized vendors, both software and hardware. Look at Red Brick and Sybase ASE on the software front. Or the old Teradata hardware line, or White Cross on the hardware side,” Madsen points out. “General purpose has won almost every battle so far.”

In the end, Madsen argues, Teradata’s future—like those of data warehousing appliance vendors Netezza Inc. and DATAllegro Inc.—lies in commodity hardware, and not proprietary (if scalable and highly available) systems. “I think [Teradata] should—and probably is—looking at how to run on commodity white-box servers, similar to what Greenplum is doing [with Sun and other vendors],” he concludes.

There’s another wrinkle here, too: NCR’s proprietary Unix flavor is probably also optimized for Teradata. One industry watcher who’s familiar with the space says that when Teradata approached vendors such as Sun Microsystems Inc. and the former Sequent about potentially supporting its data warehousing platform, these vendors declined—for this very reason.

“What I remember of the porting of [Teradata] onto other platforms is that it was a nightmare. They went to Unix vendors like Sun and Sequent and the costs were enormous. They wanted something like 400 kernel API changes to support the database. You don't change the kernel APIs on a whim.” This anonymous source concluded: “That's why these companies passed on [Teradata] support the first time around.”