Teradata: Too Rich For Your Blood?By Baselinemag | Posted 2004-02-05 Email Print
NCR's Teradata will store your databut at a steep price.
Like the penthouse at the Four Seasons or a Gulfstream IV jet, Teradata appeals
to an elite and moneyed clientele. Customers say its data warehouse, which provides
a central repository for analyzing business statistics, is peerless in its ability
to process huge amounts of informationwell beyond the capabilities of
systems from IBM or Oracle. With a Teradata warehouse, thousands of people can
simultaneously search through more than a quadrillion pieces of data without
so much as a hiccup.
"You can scale the thing as big as you want it to go," says Mark L. Andrews,
systems specialist with Federated Department Stores, parent of Bloomingdale's
and Macy's, which runs a 14-trillion-byte Teradata system on 20 servers.
That technical prowess has won Teradata, a division of NCR, a customer list of about 650 large companies and government agencies, including Albertson's, FedEx, Ford, the U.S. Postal Service and Wal-Mart.
But such rarefied technology can cost a limb or two. Teradata, which until last year didn't publicly disclose its pricing, says a standard, two-server data warehouse able to store 300 billion bytes starts at $486,000. While that includes Teradata's database software, it's as much as 10 times the price of two fully loaded, dual-processor PC servers. Teradata's software does run on generic Windows systems, but most customers opt for the company's own Intel-based Unix servers, which are specially tuned for large-scale Teradata installations.
Meanwhile, Teradata's rivals are nipping at its heels with lower-cost systems that are steadily climbing up the performance curve. "The sweet spot for Teradata is shrinking," says Gartner analyst Kevin Strange. "They're going to have to bring down pricing, or they will hit the proverbial brick wall."
Teradata insists that the notion its pricing is out of whack is more perception than reality. "This whole cost thing has to be put into the proper context," says chief marketing officer Bob Fair. First, he says, the prices of Teradata's systems are in line with those of IBM or Oracle systems that provide comparable performance. Fair says customers also need to weigh such factors as being able to create "a single version of the truth" by consolidating disparate data sources into a single warehouse. In other words, Teradata says, you get what you pay for.
The pricing issue, however, recently came to a head at one of Teradata's formerly marquee accounts: J.C. Penney, which is in the midst of moving the bulk of its four-year-old Teradata data warehouse to an IBM DB2 repository. According to several industry executives, the Plano, Texas-based retailer is making the change mainly because the Teradata system is more expensive than IBM's. J.C. Penney declined to comment to Baseline. For its part, Teradata says that J.C. Penney is still using several of its applications, and that 120 companies have migrated to its systems from other data warehouses.
All the same, customers continue to press Teradata to bring its pricing down. "The cost model needs work," says Mary-Jane Jarvis-Haig, senior manager for data warehousing at Hudson's Bay Company, a Canadian retailer. "The bill from Teradata is always higher every year."
Not everyone believes Teradata is overpriced, of course. Nationwide Insurance, based in Columbus, Ohio, has been a Teradata shop since 1998. "We still actively pursue other vendors, but they can't compare on price-performance," says Tobi Zappe, one of the firm's data-warehouse architects. In 2002, Nationwide solicited a bid from IBM for a DB2 data warehouse running on p690 servers. But Big Blue's proposal was actually $1 million more than the approximately $18 million the company had already invested in its Teradata infrastructure at the time, Zappe says.
Customers also say that the cost of owning and operating a Teradata system often is lower than that of other database systems, requiring fewer staffers to run because it handles many tasks automatically. At Continental Airlines, for example, one database administrator spends half his time managing a three-trillion-byte Teradata data warehouse that contains 80% of the company's data assets. The airline says its transaction-oriented Oracle infrastructure, by comparison, needs around three times as much staff time to maintain.
But in some cases, pricing is still Teradata's Achilles heel. BISYS, which provides information-processing services to 250 community banks, switched to IBM's DB2 in 1999 after Teradata wanted roughly $750,000 for an additional two servers, according to Bill Johnson, the company's executive vice president of operations and technology. "DB2 does require more care and feeding than Teradata, but that's a small price versus what we were paying in both maintenance and upgrade fees," he says.
Headquarters: 1700 S. Patterson Blvd., Dayton, OH 45479
Phone: (937) 445-5000
Ticker: Unit of NCR (NYSE: NCR)
Business: Data-warehousing software, servers and analysis tools, as well as related consulting services
Founded: 1979; acquired by NCR in 1991 top executives: Mike Koehler, senior vice president; Stephen Brobst, chief technology officer; Mark Hurd, who ran Teradata from 1999 to 2002, is now NCR's chief executive officer
Financials: Sales of $1.21B, operating income of $145M for 2003
Products: Database and data-processing software, which run on proprietary Unix servers as well as Windows servers. Other software products include tools for data mining, customer-relationship management and database querying.
Market Share: 6% of the $3.6 billion data warehouse management software market in 2002, according to IDC.
Competitors: IBM; Netezza; Oracle
Any database can crunch numbers. What sets Teradata's apart is its ability to break complex problems into small parts and deliver answers fast.
The Teradata system takes a query and works on different parts of each answer in parallel. By breaking the request into smaller components and working on each of them at the same time, Teradata software can probe extremely large data sets with unrivaled speed and agility. Within each chunk of work (known as a "step"), the Teradata database also processes multiple operations simultaneously, using a technique known as pipelining. This way, Teradata's software can begin a new task before the preceding one has been completed.
In practice, this parallel executionwhich was the primary design point when the database was developed 20 years agocan dramatically reduce the time it takes to process queries on very large databases. Teradata cites the results of a 2003 benchmark test created by the Transaction Processing Performance Council on a 10-trillion-byte database showing its system running 31% faster than a similarly configured IBM DB2 system.
Moreover, Teradata's parallel architecture allows it to handle huge, complex queries that can cause competing systems to seize up. "You can ask any question of the Teradata database," says Alicia Acebo, director of data warehousing at Continental Airlines. "With other databases, you can run a querybut you're not necessarily going to get an answer."
U.S. Postal Service
Mgr., Enterprise Architecture and Standards
Project: Linked previously separate sources of data in departments using a 62-server Teradata warehouse. Goal: To better manage costs.
Dir., Data Warehousing
Project: Allows 1,400 employees to use 40 applications for querying and cross- referencing three trillion pieces of data stored in a 12-server Teradata warehouse.
Hudson's Bay co.
Senior Manager, Data Warehouse
Project: Retailer consolidated inventory and sales information across four stores onto an eight-server Teradata system.
Nationwide Mutual Insurance
Data Warehouse Architect
Project: Keeps property and casualty insurance policy information on a 20-server system that processes 25,000 queries a day.
Senior Dir., I.T.
Project: Optical retailer implemented two-server data warehouse as part of a demand-forecasting application that affects inventory at 1,700 locations.
Iowa Dept. of Revenue
Project: Agency spent $11.5 million on Teradata warehouse and software, which it used to identify and collect $32 million in back taxes over three years.