Data Warehousing Appliances: The Next High-End Trend

DATAllegro joins Netezza in the data-warehousing appliance market

If one vender caters to a specific customer requirement, it’s invariably pigeonholed (rightly or wrongly) as a “niche” player. But the moment an additional competitor enters the fray, last year’s unsexy niche player can quickly be transformed into this year’s bona-fide trendsetter.

To a certain extent, this has happened in the high-end data warehousing space, where Netezza Corp. more or less jump-started the data warehouse appliance market several years ago. This month, upstart vendor DATAllegro Inc. followed in Netezza’s footsteps when it announced its own line of black-box, data-warehousing appliance solutions. Call it a high-end trend.

Of course, when Netezza first launched, the data-warehouse appliance wasn’t yet perceived as an automatic slam dunk. “We thought that if we could bring some disruptive technology to this market—that is, performance priced for lower overall cost of ownership—and present that as an appliance to the enterprise market, that would be something that could really shake up the status quo,” says Jit Saxena, Netezza’s co-founder and CEO.

Enter DATAllegro

At this month’s TDWI World Conference, DATAllegro showcased its P3000, a data-warehousing appliance that can be used by itself or bolted-on top of an existing Oracle data warehouse. The P3000 can support up to 3 TB of user data and—as high-end solutions go, anyway—is affordably priced at $450,000.

DATAllegro’s flagship data warehousing appliance uses commodity hardware, like the Netezza solutions it’s designed to compete against. A single unit can support up to 42 Intel Xeon processors, and the P3000 uses inexpensive Serial ATA storage instead of SCSI or fibre channel devices. It’s also based on open-source operating system (Linux) and database (Ingres) software. The Xeon chips that form the basis of the P3000’s power plant are outfitted with Intel’s new EMT64 (64-bit) extensions, which should effectively allow Xeon to scale to near-RISC or -Itanium heights.

The data warehousing appliance idea is now half a decade old, but, then as now, its major selling point remains ease of use. “The appliance packaging per se was really driven by the ease of use kind of issues,” says Saxena. “Now that the interface by which applications access databases has been standardized—because of JDBC, ODBC, and SQL—then the rest of the stuff could be packaged as an appliance with a plug-and-play attribute.”

Of course, it’s no coincidence that Netezza and DATAllegro have targeted the middle to high-end of the data warehousing market, either. Both companies claim substantial performance advantages relative to their conventional competitors—with the obvious exception of NCR Corp.’s Teradata subsidiary, of course. “What we’ve done is, because we’ve integrated the whole thing, we’re just getting much higher levels of throughput,” says DATAllegro CEO Stuart Frost. “When you start talking about a single appliance getting 16 GB/s of throughput, that’s just way beyond what a typical customer would do.”

But the 1 to 5 TB data warehouse market DATAllegro hopes to target with the P3000 isn’t exactly bereft of competitors. There’s Netezza, for starters, but there’s also Teradata, which, for more than a decade, has been the last word in high-end data warehousing.

“DATAllegro is entering a highly competitive market where, in addition to database generalists Microsoft, Oracle, IBM, and Sybase, it faces data warehouse specialists, including high-end leaders like Teradata, and data warehouse appliance pioneers like Netezza, a company whose efforts have served to give credibility to the data warehouse appliance concept,” notes Mike Schiff, a senior analyst with consultancy Current Analysis Inc.

In the ETL space, it’s often difficult to get any two vendors—e.g., Ascential Software Corp. and Informatica Corp.—to say anything nice about one another. In the data warehousing appliance market, however, the opposite seems to be the case—on the surface, at least.

“This space overall needs two strong competitors,” says Frost. “We have nothing but good things to say about [Netezza]. They’ve proven that the appliance model is attractive to people and sustainable, and what they do, they do very well. But our architecture is a little bit different.”

Netezza’s Saxena, for his part, positions DATAllegro’s challenge as still another vindication of the appliance model. “I don’t know much about [DATAllegro], to be honest, but I do expect that you will see more competitors in this market as we move forward, because it clearly makes sense to provide an appliance-like packaging into this market.”

Both company chiefs claim that there’s plenty of upside to the appliance market, and both vigorously rebut the suggestion that they’ll mostly be fighting over the same customers—or nibbling at Teradata from the bottom, as Netezza has done. Once you get down to brass tacks, however, much of this comity evaporates.

“We have an advantage with our architecture [over Netezza] because we designed it right from the ground up to be fault tolerant. Netezza’s not quite there in terms of fault tolerance—in fact, to get an equivalent level of fault tolerance with Netezza, you’d need to put [two appliances] together in an active/active configuration,” claims Frost.

Not surprisingly, Netezza’s Saxena disputes this, conceding that while this claim might once conceivably have been true, it’s no longer the case. “That’s a moot point for us,” he contends. “Our products—like the 8450 [and] the 8650—are all available with total redundancy, and we have always had redundancy at the data storage level, but now we have redundancy at the host level as well. So whether it’s the power supply, the networking between the processors, [or] the storage, we have this total redundancy.” But at what cost redundancy? “On certain products, it is standard, on some of the other products, it’s an option,” Saxena says.

Not Just After Teradata’s Leftovers

For years, the high-end data warehousing space has been dominated by Teradata, which markets offerings based on its own hardware, operating environment, and database. Teradata is not technically an appliance, of course, because the company does not use an open, commodity operating environment (such as Linux) or database (such as MySQL).

Since it first began shipping data warehouse appliances three years ago, Netezza has often been compared to Teradata, perhaps because it targets many of the same (smaller) customers that might otherwise opt for a Teradata solution, or maybe because Teradata itself is a “black-box” proposition of a sort. For this reason, analysts have concluded that Netezza and Teradata compete for a portion of the same market. Saxena disputes the assumption, however. “I think the bigger market for us are people that have a need for very-large-scale data analysis that actually cannot sort of go the Teradata route because of the performance and the cost of ownership reasons,” he says.

Current Analysis’ Schiff, for his part, says that Netezza has positioned itself as a “low-cost, quick implementation alternative to Teradata, especially at the departmental or functional level.” The irony, he notes, is that with the emergence of DATAllegro, Netezza may get a taste of its own medicine.

In this respect, DATAllegro’s Frost sees a built-in opportunity for a vendor willing to target the low-end of the high-end—in other words, a portion of the same bread-and-butter market serviced by Netezza. “What we really try to do is fit into the gap between Oracle and Teradata,” he says. “Oracle runs out of steam after 1 TB or so, and Teradata isn’t really interested in sites less than 5 TB. So we think it’s an underserved part of the market.”

About the Author

Stephen Swoyer is a Nashville, TN-based freelance journalist who writes about technology.