Showcasing the Best and the Brightest at Teradata Partners
Teradata touted a new Extreme Performance data warehouse appliance and its first-ever enterprise-oriented cloud offering, and that was just the beginning.
Teradata Corp. used its annual Partners confab to showcase a pair of signal deliverables: an enterprise-oriented cloud offering and a new -- though not yet available -- "Extreme Performance" data warehouse (DW) appliance.
Teradata also touted an improved geospatial facility; call-center collaboration with Microsoft Corp.; and additional progress in its long-percolating partnership with SAS Institute Inc. In educational tracks, officials also talked up Teradata's implementation of the open source "Hadoop" MapReduce API.
That was just the beginning.
Teradata bills its new SSD-based Extreme Performance appliance as the industry's fastest data warehouse. It isn't the only vendor to make that claim, of course, and with Oracle Corp.'s latest and most ambitious assault on the DW high-end -- Exadata version 2 -- competition in the DW arena is more fractious than ever. It's more tendentious than it's ever been, too. Oracle, for example, based its claim to performance bragging rights on a then-unsubstantiated TPC-C benchmark. Similarly, columnar database specialist ParAccel Inc. published (and subsequently withdrew) a TPC-H benchmark that it claimed established its position as the industry's fastest platform.
Teradata, by contrast, didn't produce a benchmark to substantiate its own claim. According to Randy Lea, vice-president of product and services marketing, it doesn't see a need to do as much. When pressed a bit harder, Lea and other officials stop just short of saying that Teradata doesn't have to.
For more than 20 years, Lea points out, Teradata's technology has delimited the high end in the DW segment. This isn't just hyperbole, either, asserts Dan Graham, senior marketing manager for Teradata's Active Warehouse efforts.
Graham cites Teradata's query-level support for mixed workloads, a differentiator he claims is matched by only two other competitors -- in very different market segments. "There's only a handful of machines in the world that can do this. The IBM mainframe has done mixed-workload management for 30 years. The Tandem machine can do it, but it's still mostly stuck in the investment [banking] space, and I can assure you that what … [Hewlett-Packard is marketing] with Neoview is not a mixed workload [solution]," he argues.
"Teradata [supports] mixed workloads at the query level. We're the only ones at the decision support-level doing this. We're on our fourth major generation. We've been doing this since the early 90's. This code core has been in there since the 1990's."
In the case of its new Extreme Performance appliance, Lea continues, the use of SSD storage -- which he explicitly contrasts with the Flash Cache modules used in Oracle's Exadata configurations -- gives Teradata superior I/O and throughput performance, relative to its existing top-end systems. (The two technologies are themselves different; what's more, both companies use them in different ways -- and to different ends -- which effectively makes comparison inapposite.)
"We're basing the [performance] claim [on the fact] that we know what our performance is against ourselves, and we're beating other people with our platform," he says. "We're beating ourselves by a large enough volume that we feel comfortable saying that we're the fastest [data warehouse in the industry]."
One caveat, of course, is that the Extreme Performance systems are still in beta, with availability slated for the first half of next year. The market isn't sitting still, however. Several of Teradata's rivals are busily prepping Flash-powered analytic appliances of their own. Netezza Inc., for example, expects to deliver a "lean" version of its TwinFin appliance systems by late next year. That appliance will feature a larger-than-standard memory complement, along with -- drum roll, please -- Flash or SSD-based storage. Teradata likes to counter that neither Netezza nor for that matter any of the other analytic database players are -- in the strictest sense -- symmetrical competitors, inasmuch as none of them markets full-fledged data warehouse platforms, complete with a constellation of connectivity, management, and similar tools.
Although that might be the case, Oracle is a symmetrical competitor and Exadata version 2 is available right now. (What's more, IBM has announced a pair of Exadata killers: its Smart Analytics black box -- a preconfigured data warehousing offering -- and DB2 pureScale, an OLTP clustering technology. Smart Analytics was announced in July; DB2 pureScale is slated to ship at some point in December.)
To Teradata's credit, it was one of the first DW players to talk up the potential of Flash and SSDs. At last year's Partners meeting, in fact, the company showcased a prototypical SSD-powered system, trumpeting -- at the same time -- an ambitious "temperature-based" storage management strategy. (In this scheme, "hot" -- or frequently-accessed -- data gets moved to super-fast Flash or SSD storage; "warm" data to fast HDDs; and "cool" data to bigger, clunkier HDDs. The Teradata database manages everything.)
In an interview last year, Lea said that SSDs weren't feasible primarily because of their high cost; 12 months on, he says, costs have come down -- big time. At the same time, Lea concedes, the Extreme Performance appliances will still be priced at a premium relative to Teradata's other systems. On the other hand, he suggests, some -- perhaps many -- customers will be willing to pay a premium for that kind of performance.
Teradata's Enterprise Analytics Cloud is -- in a certain sense -- characteristically Teradata. Over the last six months, for example, several analytic database players announced cloud computing strategies; some of these have centered on the use of public cloud services -- such as Amazon's increasingly ubiquitous Elastic Compute Cloud (EC2). The Enterprise Analytics Cloud has an EC2 use-case, too, but Teradata suggests that its most effective use case will involve private deployments alongside -- or on the same systems as -- existing Teradata deployments. Regardless, says Lea, the Enterprise Analytics Cloud permits the customer to decide -- and even has a VMWare option (via VMWare Player), too.
There's a sense in which the Enterprise Analytics Cloud brings a utility-like computing experience to Teradata environments: in one deployment scenario, a customer opts to configure a certain percentage of an existing Teradata cluster as a private Enterprise Analytics Cloud. The customer can then allocate capacity however it sees fit, spinning out Teradata instances to developers, Q&A testers, individual business units, ambitious analysts or super users, and other (oft-neglected) user constituencies.
Lea, sounding almost like a veteran cloud proponent, cites a litany of common scenarios -- namely, prototyping; quick-and-dirty deployments; seasonal or one-off workloads; and nice-to-have projects that (in the existing power structure) simply don't get greenlit -- that would be ideal candidates for deployment in a Teradata-based private cloud.
For existing Teradata customers, he concedes, a public cloud deployment on top of EC2 (or a private cloud setup with VMWare) might not make a lot of sense. For prospective customers -- particularly those that might be intrigued by Teradata's platform-centric view of enterprise data warehousing (to say nothing of its creditable analytic performance) -- these scenarios make a lot of sense: a prospect can quickly prototype a Teradata instance on either EC2 or VMWare.
Similarly, existing Teradata customers might simply want a quick-and-dirty way to expose additional analytic capacity to one or more end user constituencies, particularly for seasonal or one-off projects, without also impacting (or sapping capacity from) their existing Teradata Warehouse systems. EC2 gives them a means to do that.
Teradata had plenty of other news on tap at Partners. It teamed with Microsoft, for example, to introduce a new offering -- dubbed Teradata Contact Center Intelligence for Telecommunications -- based on Redmond's Office, Portal, BI, and call-center tools.
Elsewhere, it touted a geospatial capability that -- thanks in part to its work with SAS -- is several orders of magnitude faster than the (SAS/Teradata) status quo. The firm also trumpeted its nearly two-year-old relationship with SAS, citing a combined 350 joint customers and additional progress -- newer, deeper support for SAS user-defined functions, and improved support for SAS' "quirks" -- in terms of embedding SAS analytics in the Teradata database itself.
Teradata's support for MapReduce -- via the open-source Hadoop implementation -- isn't something it's actively promoting, says Graham.
In one educational session, Teradata CTO Stephen Brobst did discuss how Teradata has implemented MapReduce – and (moreover) how it plans to accommodate that technology in the future. It's an embryonic area, officials stress, but it's one that bears careful watching -- and not just from a Teradata perspective.
"MapReduce as it sits today is embryonic. It's clumsy. It doesn't have tools. It has a lot of excitement and momentum. There are a lot of installs, but no two installs are the same," he explains. "In the next five to seven years, these things will grow up."
Teradata hasn't officially announced support for MapReduce because it isn't "announcing a big strategy or a big product here," Graham continues. "We don't have any product slated to [support MapReduce] in the upcoming releases." Instead, Graham and Brobst say, MapReduce -- like Teradata's support for SAS workloads or its enhanced support for UDFs and non-SQL stored procedures -- expands the option sets available to customers.