In-Depth
Kognitio Stays True to Row-Based Databases
At a time when some in the industry have gone columnar crazy, Kognitio plans to stay true to what works in database technology.
In an analytic database segment that positively teems with colorful stories. It descends from the former WhiteCross Systems, which -- along with the former Red Brick Systems Inc. -- help give birth to the specialty data warehousing (DW) segment in the mid-1990s. Kognitio's stake in the current analytic database segment goes back to the first half of the first decade of the 21st century.
At about the same time that competitors Netezza Inc. and Greenplum Software Inc. were stirring things up in the U.S., Kognitio was ramping up in its bread-and-butter UK market. It remained a UK-only vendor for most of the 2000s; then, in early 2008, Kognitio made its U.S. debut. Since then, it has worked to stake out its market niche, touting a WX2 analytic database that boasts what it says are best-in-class, in-memory, and fault-tolerant capabilities, as well as creditable workload management and dynamic reconfiguration amenities.
Kognitio officials offer a split take on the analytic database market.
On the one hand, says U.S. CEO John Thompson, near-non-stop innovation on the hardware side of the market everyone -- Netezza, Kognitio, Aster Data Systems, even Teradata Corp. -- now competes on a level hardware playing field. On the other hand, Thompson says, hardware parity means that vendors must increasingly compete on the basis of software differentiation. (See http://tdwi.org/Articles/2010/03/24/Scale-Processor-Analytics.aspx?admgarea=news&Page=4.)
In this respect, he maintains, Kognitio has an especially compelling story to tell.
"When you're looking at some of the older players in the marketplace, and we don't cast aspersions for maturity per se, because we're clearly 20 years in the making ourselves -- but [with] some of the older players, you can't really do real-time continuous loading into the database," Thompson observes. "That's a make-or-break feature. We've seen the complete collapse of the batch window, and we've seen a collapse [of the status quo] in the way that databases are looked at. Today, most of the people who are building business intelligence and data warehousing infrastructures assume that they are operational and have to be available pretty close to 24x7."
Real-time or near-real-time isn't for every enterprise, but in the Big Data markets in which Kognitio and other analytic database vendors compete, Thompson says, it's becoming a de rigueur requirement.
"What we've seen is that any time that we're talking to any of the online players -- say, some of the newer players in the marketplace, [such as] social media people or the community-based couponing [vendors] -- any time you talk to any of those organizations, it's all about real-time, any-time, all-the-time data warehousing," he says. "It's the same with financial services: they're all about fast fast fast and always-on as well. It's something you have to be able to [address] if you want to compete in this space."
Kognitio's U.S. introduction preceded by about a year the Second Wave of analytic database players -- i.e., the arrival of columnar rivals such as Aster Data Systems Inc., Infobright, ParAccel Inc., and Vertica Inc.
Unlike these players, Kognitio's WX2 database isn't a column-based repository, nor is it based on an existing free and/or open source software (F/OSS) DBMS. Like Vertica, WX2 is homegrown; unlike Vertica -- which likes to pitch its flagship analytic database as a brand-spanking-new DBMS deliverable -- WX2 is almost two decades old.
Thompson, not surprisingly, positions WX2's longevity -- its "maturity" -- as a differentiator. For one thing, he points out, there's nothing new about the column-oriented DBMS. In fact, he argues, column-based stores such as Sybase IQ had been consigned to a specific and limited market niche for decades.
The columnar renaissance -- (or, from the perspective of columnar detractors like Kognitio or Dataupia) columnar resuscitation -- is a product of the last five years. In this respect, Thompson claims, Vertica and other neo-columnar databases developed specifically in response to analytic workloads that were overwhelming consumer, off-the-shelf (COTS) DBMSes. Those issues are now "receding," he contends.
"My view is that columnar is a really interesting and good technology for certain applications, but … I believe that those applications are receding and becoming more and more of a minority in the trend that we see going toward Big Data [and] Always-On Data," he maintains.
Thompson, like other columnar skeptics, raises questions about the general desirability of a column-oriented architecture, at least from a data management perspective. This isn't surprising: columnar haves (such as Aster Data, Infobright, ParAccel, and Vertica) tend to champion column-orientation as a must-have for Big Data problems of scale (see http://esj.com/Articles/2010/03/02/Data-Management-Crossroads.aspx). Those not in the columnar camp (such as Kognitio and Dataupia) tend to concede advantages in some (very specific) cases, but inevitably raise questions about column-orientation's suitability in "broader" or "general" DW scenarios (see http://tdwi.org/articles/2010/03/17/dataupia-hinshaw-back.aspx). Vendors such as Netezza and Greenplum -- which recently introduced hybrid row/column facilities for their DBMSes -- tend to be more pragmatic on both questions (see http://tdwi.org/articles/2009/12/16/bi-the-year-in-review.aspx).
"The drawback with columnar is that if you're constantly loading data and you want to give your users the ability to ask anytime/anywhere questions … that's going to negate" columnar's purported advantages, Thompson argues. "If you have applications where people are asking pretty much the same questions over and over again, you might have the time to do that, but if you're all about advanced analytics, where you're empowering users to [bang away at] the warehouse with all kinds of custom or seemingly random queries, that isn't ideal."
This, too, is a somewhat familiar line of attack. If it amounts to an oversimplification of the issue -- an accusation that (for the record) Thompson rejects -- so, too, he asserts, does the typical columnar take on row-based DBMSes such as WX2. "This is exactly [the market segment] where Kognitio and WX2 have lived for the last 20 years," he avers.
"We enable the analytically-oriented people in your business to start at one point and traverse the structures or hierarchies [of the database] to come up with the answers they need for their business. To claim that [columnar] targets problems that weren't being addressed by any existing [technologies] is just wrong."
The real "stuff" of advanced analytics involves asking new or unorthodox questions, Thompson continues. "The real value, the real payback in these [analytic] infrastructures, is to ask the questions that weren't asked before. For example, the questions you ask on Monday might not be the questions you ask on Wednesday, or on Friday, either. The point is to empower users to answer new and innovative questions that drive business returns."