Need further proof that competition in the data warehouse market is as hot as ever? How about this: Not one but two vendors are giving away their analytic databases free of charge, albeit with some strings attached.
First up, Greenplum. The San Mateo, Calif.-based vendor last month introduced a free, single-node version of its flagship massive parallel processing (MPP) data warehouse aimed at data analysts "working on the fringes" of more traditional enterprise data warehouse deployments.
Not to be outdone, ParAccel, which was founded in 2005 by execs from database heavyweights Oracle, Teradata and Netezza, is offering customers a year's worth of software licensing fees gratis if they trade in their current data warehouse appliances.
Whether innovative marketing campaigns or desperate gimmicks to drum up new business, both moves illustrate the lengths vendors will go to in order to differentiate themselves in a market where prices are coming down even as performance goes up.
"It's not just a price war, it's a performance war," said Jim Kobielus, an analyst with Cambridge, Mass.-based Forrester Research.
In addition to the advent of MPP and columnar-based architectures, the improvement in the price-to-performance ratio of data warehouses is due largely to improvements to the commodity hardware that powers them.
Commodity hardware from the likes of IBM, HP and Sun, for example, allow vendors to tune their data warehouse appliances for high performance and scalability and at the same time keep prices low, Kobielus said.
ParAccel in particular is trying to exploit this trend. Its "Cash for Clunkers" program, as the vendor is calling it, is based on the premise that customers want choice when it comes to which hardware powers their data warehouse appliances. Customers, presumably, will opt for the cheapest hardware.
Earlier versions of appliances from Oracle, Greenplum and Netezza run on only specific hardware, ParAccel claims, locking-in customers and depriving them of the benefits of commodity servers and storage. ParAccel says its data warehouse appliance can run on any of a number of vendors' hardware, giving customers the choice they want.
"It's that hardware choice that people really want and [they] are limited still by the first-generation appliances," said Kim Stanick, vice president for marketing at ParAccel. "As soon as you get locked in to a situation where you don't have choice, you now limit your ability to take advantage of that increase in price/performance."
ParAccel's basic premise -- that most data warehouse vendors don't offer hardware choice -- could be flawed, however, or could at least quickly become outdated, according to Kobielus.
Vendors like Netezza and Teradata have indeed highly tuned their analytic database software for specific hardware for maximum performance and scalability, he said, and "some customers want that optimization."
But most major data warehouse vendors have little choice but to adopt a hardware-agnostic approach in order to stay competitive and keep their prices below the $20,000 per terabyte threshold, he said.
"The DW appliance companies themselves recognize they cannot afford to be over a barrel to one hardware vendor as well," Kobielus said. "Lock-in is something that everybody [not just ParAccel] is moving away from at various speeds."
When it debuts next year, for example, Microsoft's SQL Server 2008 R2 parallel data warehouse is likely to leverage commodity hardware from multiple vendors, Kobielus said. Microsoft recently even told analysts that the starting price for the database, code-named Madison and built on DATAllegro technology, will be as low as $13,000 per usable terabyte.
Netezza's TwinFin data warehouse appliance, which debuted in August, also takes advantage of commodity hardware.
While also offering a free version of its database, Greenplum is taking a tack different from ParAccel's in order to lure new customers. Greenplum envisions a "distributed model" of data warehousing where data analysts can integrate data from multiple sources into smaller data marts for customized analysis, rather than being forced to work within the confines of an enterprise data warehouse.
Greenplum Database -- Single Node Edition, in conjunction with Greenplum's Enterprise Data Cloud initiative, is designed to let analysts quickly create and share new, small data marts free of charge, according Ben Werther, the vendor's head of marketing.
"You can do a lot of analysis using eight or more cores of processing on a single node," Werther said. "You'd pay vendors like Oracle a lot of money for that, and in this case you can do that at no cost."
Werther also addressed the issue of hardware lock-in, noting that earlier versions of Greenplum's database were highly dependent on Sun hardware. But with more recent releases, the vendor has embraced competing hardware from IBM and HP. Deals including Sun hardware now account for only about 10% of Greenplum's sales, he said.
"We're really building platforms on all the major Tier-1 [hardware] vendors and offering customers a lot of choice on how they deploy us," Werther said.
Whether gimmicks or not, deals like the ones on offer from ParAccel and Greenplum make clear that the battle for data warehouse supremacy is still being waged. "They have to do this now," Kobielus said. "There's just so much competition in this market."