In-Depth

DATAllegro Ups Ante in DW Appliance Segment

Whatever else it might be, the data warehouse appliance market is anything but sleepy.

• By Stephen Swoyer
• 08/22/2007

Whatever else it might be, the data warehouse (DW) appliance market is anything but sleepy. From quasi-appliance giant Teradata’s impending split from parent company NCR Corp., to the entry of new appliance players (such as Dataupia, which launched in May), to the growing presence of established players (such as GreenPlum), the DW appliance segment teems.

Things got even more interesting this week, when "veteran" DATAllegro Corp. announced a couple of new wrinkles on the tried-and-true DW appliance model. First, DATAllegro unveiled a new "low-cost" offering designed to address data archiving scenarios. "Low-cost," in this context, anyway, means about $1.2 million dollars—for a rack (approximately 200 TB) of user data.

Secondly, DATAllegro announced a grid-ready DW appliance. Grids aren’t quite the silver bullet some proponents have made them out to be, but DATAllegro CEO Stuart Frost says there are situations in which computational grids—predicated on the familiar hub-and-spoke paradigm—can help accelerate certain applications. "People … wanted us to do more [in the area of] mixed query workloads, so more and more concurrency," he comments. "What they were asking for was [a capability] more and more like Teradata—but cheaper. We were starting to think whether there was another way of doing it that was maybe easier from our point of view on an R&D basis, and so we came up with this."

DATAllegro’s grid support lets it connect separate appliances node-to-node in what is effectively a hub-and-spoke grid. "We can do parallel data transfers between [nodes], federated queries across them, all at a very high speed," he explains. "Hub and spoke is really the main use of grid technology that we see. It really does fit into that large enterprise use of data, [which is to] push [data] into data marts or appliances. [These customers are] concerned about the lack of governance, they’re concerned bout SILO-ing, but until now they didn’t have anything that could handle [such query requirements at a low price point]. So a lot of people recognize that hub and spoke is the right way to go for that."

DATAllegro’s massive new user data racks focus on a different kind of problem, Frost indicates. "We’ve got customers where right now they’ve got a high speed appliance with the last six months of data. Now what they want is something that they can easily query over the last two years [of data]," he explains. "We can give them up to 200 TB of user data in a single rack, lifting in one of those [racks] at $1.2 million."

The attraction, he argues, is that customers can enjoy simultaneous access to fresh (or mostly) current data, as well as rapid access to historical data: "You can connect the high speed appliance and the larger archive appliance together to do high speed data transfers between them, so as new data is loaded it can be rolled off to the larger rack. And this can be automated. Queries can just automatically be federated between the two appliances, too."

A lot has changed in the DW appliance segment since DATAllegro announced its switch to commodity servers, switches, and storage six months ago. (http://www.esj.com/business_intelligence/article.aspx?EditorialsID=8334) Rival Teradata, for example, is currently in the process of splitting off from its parent company, NCR. And DATAllegro arch-rival Netezza Inc.—which more or less created the DW appliance category—is prepping for its initial public offering.

What’s more, still another DW appliance player—Dataupia, the brainchild of several former Netezza backers—has also crashed the party. But DATAllegro’s Frost is as sanguine as ever about both his company’s prospects and the outlook for DW appliances in general. "We actually sold one of [our grid appliances] a couple of months ago, and we’ve got a customer who’s about to go live on a 320 TB [configuration] in just a few weeks," he points out. "We’ve been pretty successful, too, since we switched over [to Dell servers and EMC storage], and most of our deals are 50 TB and up."

While DATAllegro and Netezza have traditionally attacked Teradata for being too expensive, Teradata has countered by trying to pigeonhole both vendors—and the data warehouse appliance in particular—as a band-aid solution of sorts. (http://www.tdwi.org/Publications/display.aspx?Id=7799) In other words, Teradata officials like to argue, DW appliances work great as data marts—which is how most customers deploy them. But appliances tend to run out of gas—and quickly—when they’re asked to scale to support enterprise data warehouse (EDW) deployments, Teradata officials argue. (http://www.esj.com/business_intelligence/article.aspx?EditorialsID=7476)

Frost counters: "A data mart is a little bit in the eyes of the beholder. If you’ve got 320 TB in there and you’re running multiple applications from it, it may not be an enterprise-wide data warehouse in the traditional sense of the term, but it’s certainly very strategic to [the customer], and they are definitely running multiple business applications from it," he argues. "Is that an enterprise data warehouse? Yeah. I would say so. It’s enterprise in scope and scale; it’s probably one of the largest data warehouses ever created."

In the past, Frost concedes, DW appliances did tend to find traction largely as turnkey data marts. "In some ways, this is correct. Clearly, ourselves and Netezza have sold into business units and have sold data marts, but really this whole strategy [with DATAllegro’s grid and archive solutions] is designed to move into the next level of that," he argues.

Finally, Frost says there’s plenty of pie to divvy up in the increasingly fractious DW appliance segment—even with new players like Dataupia, and established stalwarts like GreenPlum, increasingly making their presences felt.

"It’s certainly a very big market. We haven’t seen Dataupia competitively yet. I think they’re playing in a different area of the market than us. They tend to go into the Oracle sites, and really be viewed as an accelerator for certain query types," he concludes. "We tried that when we started out, but we were finding it quite difficult to federate the queries across Oracle and us. I don’t know how Dataupia is going to solve that. We see GreenPlum sometimes now, too. It’s down to around three to four percent of the [customer] opportunities, whereas Netezza is more like 50 to 60 percent [of competitive situations]. And we’re seeing more and more of Teradata, too."