In-Depth

What Defines a DW Appliance?

What, exactly, is a data warehousing appliance: hardware, software, or a combination of the two?

• By Stephen Swoyer
• 08/06/2008

Last month's acquisition of DATAllegro Corp. by Microsoft Corp. raised many questions. Some, such as "What are Redmond's precise plans in the data warehouse (DW) appliance segment?" probably won't be answered for a few months yet, which is par for the course in any acquisition.

Microsoft's move also raised questions about the DW appliance segment itself. For example, when is an appliance truly an appliance? At a time when many vendors -- including, to a degree, DATAllegro itself -- tout DW appliance software that is decoupled from (or, at least, not inextricably married to) an underlying hardware platform, it's an intriguing question.

DATAllegro -- with its February, 2007 announcement of a range of appliances based on Dell Computer Corp. servers, EMC Corp. storage, and Cisco Systems Inc. networking interconnects -- seemed to sever its appliance line from any explicit dependency on hardware. DATAllegro's "appliances" still require dedicated device drivers, experts say, but can run across a range of off-the-shelf hardware configurations. DATAllegro, for example, has since added systems from EU hardware stalwart Bull (developer of the NovaScale line of servers) to its appliance line.

It's a point that DATAllegro officials were at pains to make last month, talking up the importance of DW "reference architectures" -- in which DW software is certified to run on top of a range of different hardware configurations -- to RDBMS giants such as Microsoft and Oracle Corp. (see http://esj.com/business_intelligence/article.aspx?EditorialsID=9056).

"Reference architectures are the way major vendors have co-opted the appliance story. A big part of why [Microsoft] bought us is that we have the expertise and the ability to take that [reference architecture] to a whole new level and leapfrog Oracle in [terms of] getting that reference architecture right," said Frost.

It's a question-begging point, however, according to veteran data warehouse architect Mark Madsen: in what sense, after all, is DATAllegro still an appliance proposition? "[T]hey really weren't an appliance vendor, since they just provided software. The EMC/Dell deal was the transition where they still sold it as an appliance, but there wasn't really a requirement to run it on that hardware," argues Madsen, a principal with DW consultancy Third Nature Inc.

If DATAllegro -- unlike Netezza, for example, which taps Power chips from IBM to accelerate its Snippet Process Units (SPU) -- isn't married to any specific hardware architecture, what does its software actually do? Madsen, for his part, isn't sure: "If they layer [their software] completely on top of the database, then there isn't much there because to get real performance and stability you still have to muck around somewhat lower than the SQL processing layer."

It's a question that could easily be directed to many of DATAllegro's appliance competitors -- companies such as ParAccel Inc., Vertica Systems Inc., InfoBright Inc., or even Kognitio. These vendors tout "appliance" solutions that -- simply by virtue of the availability of separate DW software -- aren't wedded to any specific hardware platform.

In other words, if you don't want to buy the whole appliance kit and kaboodle from ParAccel or Kognitio, they'll be happy to sell you just their software, which you can deploy on top of new or existing hardware assets.

There's clearly a sense in which the term "appliance" is being misappropriated, says Philip Russom, a senior manager with TDWI Research. In DATAllegro's case -- i.e., inasmuch as that vendor doesn't separately sell itself DW software -- it's probably more of an "appliance," per se, than competitive offerings. Like it or not, writes Russom in a recent article in TDWI's members-only FlashPoint newsletter ("Redefining the Data Warehouse Appliance"), the DW appliance concept is veering away from the category-defining systems first developed five years ago by Netezza.

"[T]he term 'data warehouse appliance' has been misappropriated by several vendors, from small ones with oxymoronic software appliances to large ones that offer bundles of hardware and software products," Russom says. "Although we may not like it, the term 'data warehouse appliance' means many things nowadays, and the original meaning [popularized by Netezza] has eroded considerably," Russom continues. "In that context, I'd say that DATAllegro comes closer to the original meaning than most."

Whether it's an appliance or not-exactly-an-appliance, today's "appliance-like" products all address similar needs, argues John Santaferro, director of business intelligence portfolio marketing with Hewlett-Packard Co. (HP). Santaferro can credibly claim to know whereof he speaks: HP not only markets its own dedicated appliance (NeoView), but also touts an appliance-like offering -- in this case, the HP BladeSystem for Oracle Optimized Warehouse (OOW) that it developed in tandem with Oracle Corp. In both cases, Santaferro argues, HP touts an appliance-like value proposition that addresses similar customer needs.

"Whether you're looking at an appliance or an appliance-like offering, you're still addressing the same problem, and in degrees," he comments. "The beauty of [something like OOW] might not be that it's an appliance -- because the word 'appliance' sets a level of expectation with the customer, and historically, in the IT world, … [of] something that you plug it in, it's in a box, it's closed off, there's a shell built around the hardware, so that when you plug it in, it just does its thing with little setup involved," Santaferro says. "A lot of the folks who call themselves an appliance, I'm not sure they've achieved that with their products."

What these products do achieve, he says, is a degree of plug-and-play-ability.

"The promise of [an appliance such as] NeoView or an appliance-like [product] such as the Oracle Optimized Warehouse is that you're bringing everything into this consolidated environment. You're making it much easier [for customers] to focus just on solving business problems, instead of focusing on the hardware, or the tuning, or trouble-shooting any issues that are bound to crop up," he concludes.