Vertica Enhances Flagship Analytic Database
Vertica says the latest revision of its analytic database is easier to deploy, manage, and scale.
- By Stephen Swoyer
- June 9, 2010
Some analytic database players like to emphasize their longevity. Vertica Inc., which late last month unveiled version 4.0 of its flagship Vertica analytic database, isn't one of them.
Where grizzled analytic database competitors such as Netezza Inc. and Kognitio (and even the former Sybase Inc.) argue that longevity translates into maturity, Vertica counters with a perspective of its own: maturity, particularly in the technological domain, comes at a cost.
From Vertica's point of view, most "mature" analytic database offerings could stand to shed a few extra pounds. Vertica, officials claim, has traditionally offered a lean, mean, and even green take on the analytic database. If the Vertica database lacks some of the bells and whistles of more established competitors, it likewise lacks features that encumber traditional database designs. Vertica says its database avoids some of the problems -- architectural decisions or design compromises; retrofittings; kludges; a need for backward compatibility in the face of evolving standards -- that can constrain both the performance and the reliability of "mature" databases.
"We wrote this product from the ground up. The first 20 customers were all the people who were willing to get in and deal with the warts and do everything at a command-line prompt. Because of the speed, they were willing to make that investment. [With] the next 100 customers, [what we heard] was, we had to have the GUI environment [to be competitive]," explained Dave Menninger, vice president of marketing with Vertica, in an interview in March.
Vertica, unlike many of its analytic database competitors, doesn't shy away from naming (customer) names. To date, it boasts more than 50 customer references and claims more than 100 customers in all. As Vertica has looked to add new accounts and expand into new markets, Menninger concedes, customers have asked to add a feature features or amenities.
"[In the process of] expanding beyond 120 customers, … they told us, 'You have to make this dead easy to use.' They wanted it so [that] nobody's prevented from … managing the system because they don't know the command-line [syntax] or [the GUI isn't] intuitive enough," he explains.
In this update, Menninger stresses, Vertica isn't just tacking on frills. Vertica 4.0 ships with a simplified management experience; an improved connectivity feature set (including additional enhancements to its trickle load capability); and a raft of performance tweaks, including better performance in mixed workload scenarios.
The upshot, Menninger maintains, is that it should be easier to deploy, manage, and scale Vertica.
This wasn't always the case, he concedes. When it debuted, Vertica was something of a geek's dream. The company's initial demos -- which included in-booth demos at industry trade shows -- consisted of LCDs displaying command shells flanked by rack-mounted hardware.
As Vertica evolved its GUI, it likewise exposed many options, settings, or switches via the GUI. In its version 4.0 release, Menninger explains, Vertica strove for balance, delivering new features that help automate some of the more onerous, tedious, or confusing aspects of data warehouse configuration and scaling. Vertica's new "database designer," for example, promises to help accelerate data warehouse deployment and configuration -- regardless of the vagaries of underlying hardware or storage. Although geeky DBAs can still have as much command-line fun as they want, the less geeky will find that "there are far fewer knobs in the system now."
For example, Menninger says, instead of requiring geek-level familiarity with storage structures, Vertica 4.0 takes a more abstract approach: its new "load-and-query" designer promises to automatically generate optimized storage structures. "The fundamental shift was moving away from [a need to have] a deep knowledge of the [underlying] storage structures to … [having a configuration] where the number of CPU cores is what drives the optimization of the machine," Menninger explains. "That's as opposed to … the user having to have some understanding of the physical resource structures of the machine."
At some point during an update briefing -- Vertica tries to deliver two major product refreshes per year -- officials usually end up reciting a laundry list of new support for SQL features or extensions. That's the case in Vertica 4.0, too.
However, the company is tackling more sophisticated amenities, too, such as improved support for time-series analysis, which Menninger says is especially useful in analyzing clickstream data. "[Customers] typically do their analyses in discrete time intervals -- say, every few seconds -- [and] they want to compare the values in [one] time period with [values in] another time period and have algorithms based on these fixed time windows." The rub, he points out, is that such an approach has inherent gaps.
"You want to fill in the gaps [in the series] so that your algorithms can work properly. Your alternative would be to write some complex SQL involving a lot of complex joins. We created this ability to fill in the gaps," he says.
It's illustrative of Vertica's maturation: with each new release, the company still dots new SQL-1999 "I's" and crosses new SQL-2003 "Ts"; increasingly, however, Vertica focuses on more ambitious scenarios, too. It's all part of the plan, says Menninger: "Our approach was to [build this] from the ground-up: take something [that was] fast and over time apply it to more and more situations."