Analytics, DBMS News Offer Fresh Insight into Vertica
A pair of recent announcements offers fresh -- and refreshing -- insight into the new Vertica four months after its acquisition by HP.
- By Stephen Swoyer
- July 6, 2011
Vertica -- a Hewlett-Packard Co. (HP) Company -- recently announced a new release of its database platform, Vertica version 5.0. The announcement comes just a fortnight after HP unveiled its first Vertica-branded appliance, the HP Vertica Analytics System.
Taken together, the two announcements offer fresh insight into the new Vertica, a little over four months after its acquisition by HP.
Vertica Inc. started out as an analytic database provider. It touted both its first-rate pedigree (it was co-founded by data warehousing luminary Michael Stonebraker) and a database that -- according to officials -- had been architected from the ground-up for analytics. Unlike competitive offerings, which were typically built on existing or open source database technology (such as MySQL or Postgres), Vertica didn’t have the legacy baggage of competitive offerings, officials maintained. They’d likewise concede that it lacked some of the features of competitive products, too: through its first few releases, for example, Vertica continued to augment its DBMS power with additional SQL functionality and a bevy of other missing features.
Vertica could also lay claim to another kind of uniqueness: namely, a surfeit of customers. In an analytic database arena that sometimes seemed starved for customer references, Vertica touted more than 100 on-the-record customers. It was perhaps for this reason that HP came knocking back in February.
The Vertica Analytics System that Hewlett-Packard unveiled is pretty much what you’d expect, given HP’s status as a systems giant: it consists of three different configurations (i.e., quarter-, half-, and full-rack capacities) running on HP hardware. HP positions the new Vertica Analytics System as part of its line of Converged Infrastructure solutions -- i.e., pre-configured systems comprising server, storage, networking, and software components.
The idea is that customers aren’t purchasing an appliance but an infrastructure solution. For some shops -- particularly for one on the prowl for a forklift DW program -- this kind of package could be attractive. It’s a scheme that’s been adopted, in a variety of guises (and to a variety of degrees), by competitors such as EMC Corp., IBM Corp., Oracle Corp., and SAP AG, among others.
Vertica was always sold as an either/or proposition, however. You could either buy the DBMS software and run it on the hardware of your choice or you could purchase preconfigured/pre-loaded systems from Vertica and its hardware partners. This hasn’t changed and (likely) isn’t going to change, officials stress. That’s perhaps the most refreshing take on the post-acquisition Vertica.
“[T]he plan is to continue to offer the Vertica Analytics Platform software on a number of different platforms. Right now, only about half of our customers are installed on HP servers,” says Chuck Smith, corporate communications manager with Vertica-HP. Now, as ever, Smith avers, Vertica’s software “can be loaded on any industry-standard hardware ... on virtual machines or on the cloud.”
Smith and other officials say Vertica 5.0 includes enhancements that promise to make it easier to provision and deploy Vertica on both HP and non-HP hardware -- as well as in highly virtualized or cloud environments.
“We have a cluster-cloning capability where you can take any size cluster, say 200 nodes, [and you could] hit a button and Vertica will shoot over all of the data and will fire up another cluster -- even if that [cluster] is two, three, or ten times the size,” says Colin Mahony, vice president of product and business development with Vertica.
This makes it easier to provision extra capacity -- such as quarter- or half-rack Vertica Analytic System modules -- as needed. “We’ve made the product even more elastic so that when you add a node, the data gets moved and the node comes online about 100 percent faster than it did before.”
No-SQL Lingua Franca
No-SQL is the Newest Big Thing, and Vertica 5.0 boasts a new No-SQL capability via its new software development kit (SDK).
Although MapReduce is the most prominent of the No-SQL alternatives, the promise of something like Vertica’s new SDK is that it permits a programmer to write code in the procedural language(s) of his/her choice.
In this respect, the new SDK is similar to features touted by the former Aster Data Systems Inc. (now part of Teradata Inc.), Microsoft Corp., Oracle, and Greenplum Software Inc. (now part of EMC).
The new SDK, says Mahoney, supports C and C++, with support for other languages in the works. “Eventually, we’ll support a lot of other languages [such as] Perl, PHP, and Java. We started with C and C++ because all of our customers are interested in performance and because our engine was written in C and C++,” he explains.
Mahoney says Vertica’s new SDK brings a radical kind of self-serviceability to database management.
“I used to have these conversations with customers where they’d say, ‘There’s this hashing function that’s really important to me and I really want Vertica to be able to support it.’ I used to tell them, ‘You know what, we’ll prioritize it as best we can and we’ll try to get it into our next server release.’ Now I can say, ‘You can actually write that hashing function on your own and get it in there or we can write that hashing function and give it to you -- and it has nothing to do with the [upcoming] server release.”
In addition to its No-SQL SDK, Vertica 5.0 also delivers new in-database analytics, including a new geospatial library, improved support for linear regression functions, and a new event series pattern-matching capability.
Mum’s the World on Neoview
Vertica is HP’s latest-and-greatest analytic database gambit. However, last year at this time, HP was still pushing Neoview, a fault-tolerant (FT) database platform that it inherited -- and subsequently retrofitted for data warehousing -- from FT computing specialist Tandem Computer Corp.
Just months before it acquired Vertica, HP pulled the plug on Neoview.
In a late-February interview with BI This Week, HP officials demurred when asked if Vertica would be positioned as a migration platform for Neoview customers. Four months later, Mahony likewise demurs, saying that Vertica approaches Neoview the same way it approaches other DBMS platforms.
“We work with [Neoview] the same way we work with a lot of other accounts. In fact, we have a fair number of Teradata customers that we’re working with as well,” he points out, adding that “[HP] wanted us to focus on what Vertica can do, as opposed to worrying about any other legacy product, whether it’s owned by HP or not.”
Has the Vertica business been approached by Neoview customers, either on their own or at HP’s instigation?
Mahoney again demurs. “I don’t think it [i.e., communication or coordination] has been that strong or direct.”