Teradata is a company that has come up frequently through my various case studies in a past life. Data mining customers include Discover Financial Services, Lowe's, Harrah's Entertainment, Wells Fargo and a host of other large enterprises looking to mine customer data. Medicare and Medicaid also recently upgraded its Teradata data warehouse. The momentum was evident in Teradata's first quarter (transcript): The company handily topped Wall Street estimates with earnings of $45 million, or 26 cents a share, on revenue of $367 million.
We recently spoke with Randy Lea, Teradata's vice president of product and services marketing, to get a feel for having a bulls-eye on your back.On competitors bringing up Teradata. Lea said the company is increasingly being targeted because it's viewed as a data warehouse leader. HP is hoping Neoview can take customers from Teradata and Oracle's Exadata database appliance is looking to do the same. One big rival, however, isn't calling out Teradata: That's IBM. Why is everyone pitching an optimized stack? Lea noted that Teradata has been focusing on an integrated data warehouse stack that includes hardware, software and an optimized database, I/O and memory. "But the real value is that we're 100 percent focused on data warehousing," said Lea, who noted that Teradata is rarely an incumbent on an account (typically that's Oracle).
Lea added that the requirements between a data warehousing database and a transaction one are "two different worlds." He said that Teradata has set up its software so it can manage multiple workloads without an army of database administrators.Can Teradata compete with Oracle on the database front with a proprietary set-up? The Teradata database "is no more proprietary than SQL, Oracle," said Lea. "It's our database developed from scratch. We had an idea to develop all of the execution in parallel. It's not transaction based." Where does the hardware fit in? Teradata uses Intel processors, partners with EMC and has standard Ethernet connections for certain products. Lea said that the software is the real differentiator, but the connections to cluster a series of data warehousing appliances remains proprietary. The proprietary connections are needed when there are multiple Teradata units used in parallel.
On storage, Lea said that Teradata will unveil virtual storage management tools for multiple storage devices. Teradata's approach, which uses multiple drive technologies, sounds similar to what EMC has planned. The general idea: Mix and match storage technologies to optimize for performance can costs.The competitive landscape. When asked what rivals Teradata competes with, Lea broke it down this way:
- Larger enterprise accounts: IBM and Oracle. "It has been that way for the last 15 years," said Lea.
- Department level deployments: Netezza, Sybase.
- Smaller companies: Netezza.
Oracle has a good transactional database to do things like record updates for ERP. It is challenged with full table scans and mixed workloads. In data warehousing the average customer has 30 applications. One person may be doing deep analytics on the history of a customer, another sales comparisons for the week month and year and an automated query to generate a Web site response time for a personal ad delivery. If the database is storing two years of history it may be used 15 to 30 different ways.
Lea maintains that Oracle's Exadata product was really about integrating the storage subsystem, software and an integrated disk to do full scans better. The biggest question mark is what happens to Exadata, an HP partnership, when Oracle closes its Sun acquisition.Where Teradata fits. Lea described Teradata as the layer that does the crunching underneath the business intelligence tools. It partners with the likes of IBM Cognos, Microstrategy, Oracle Hyperion, SAS and SAP's Business Objects. How's demand? Lea said demand has been pretty steady. "Companies the use analytics during bad times survive well. You can make a lot of mistakes in the good times, but in bad times you'll go out of business. Everyone is scrambling for data analytics to find their most profitable customers," he said.
Larry Dignan is Editor in Chief of ZDNet and Editorial Director of TechRepublic.