Teradata Unveils Appliance, Data Warehouse Array Technology
At its Partners user conference, Teradata announced a new appliance model, showcased the next rev of its database, and hinted at a future SSD appliance
- By Stephen Swoyer
- October 22, 2008
At last month's OracleWorld, Oracle Corp. unveiled its latest and grandest BI deliverable to date -- its Database Machine data warehouse. Next came Microsoft Corp.'s second annual BI Conference, at which the company trumpeted both its BI aspirations and its still-gestating very large data warehousing (VLDW) ambitions.
All the same, many BI and DW pros were looking ahead to last week, when VLDW king Teradata Corp., convened PARTNERS, its annual user conference. Teradata's event might've been missing the excitement of a Database Machine demo or details of a (promised) VLDW flavor of SQL Server, but it certainly didn't lack for important announcements.
Teradata officials touted a new DW appliance (priced to move starting at $16,500 per/TB), announced a new "Petabyte Club," showcased a still-gestating solid state drive (SSD) DW system, and briefed attendees on its forthcoming Teradata 13 release, slated to ship early next year.
Teradata's annual user event isn't just a popular destination for customers, either: industry leaders typically make the trip, too. One prominent DW architect --- who attends a dozen or more such shows each year but wished to remain anonymous for this article -- said that the Teradata event is "the best BI event in the industry. [There's] much more innovative stuff there than any of the others I've been to. I always learn a lot and get plenty of ideas."
Teradata used this year's PARTNERS as a forum to tout its much-talked-about plunge into the DW appliance space. Teradata officially took the appliance leap this spring, when it announced a trio of new systems -- the Teradata 550, 2500, and 5550 appliances. Since then, it has announced two additional systems: the 2550 (last month), and the new Extreme Data Appliance 1550 (last week).
For a long time, "appliance" was a word that Teradata officials refrained from using. No more. Reached last week, vice president of product and services marketing Randy Lea frankly acknowledged that -- by downplaying the merits of DW appliances (and, moreover, by rejecting its own appliance-like attributes) -- Teradata effectively priced itself out of some key markets.
Teradata's initial appliance deliverables helped make Teradata competitive in segments targeted by appliance competitors Netezza Inc., Dataupia, Kognitio, and others, Lea maintains. Its most recent appliance deliverables, on the other hand, are part of Teradata's effort to make its own mark in the appliance space.
Consider Teradata's new "Extreme Data" appliance, the 1550, which -- from a starting price of $16,500 per TB -- can scale into the petabyte range.
"The Extreme Data appliance is [designed for] a different, unique workload. A lot of our customers in the commercial media and entertainment space have terabytes and even petabytes of data. This is data that isn't usually integrated into the data warehouse, so they're not looking for the integration and, quite frankly, the higher cost of a data warehouse. They're looking for concurrency, so they can take a look at volumes and volumes of data," he explains.
"For our Extreme appliance, the core configuration is 50 terabytes, with compression, and we can scale that up to 50 petabytes, so instead of [having this data in] offline storage, you have almost instant access. It might take me half a day to load [this] data, but that's okay. All I need to run are 50 queries a day to take a look at a behavioral profile, for example."
Join the Club
Also last week at PARTNERS, Teradata touted a new "Petabyte Power Players Club" -- a quintet of Teradata customers who maintain DW configurations of 1 PB or larger. Not surprisingly, there's an appliance angle here, too. Thanks to its new appliance deliverables, Lea argues, Teradata can scale to address the needs of entry level -- i.e., enterprises with a terabyte or less of data -- to petabyte-class customers.
"If you think about it, we have become a one-product shop for the most sophisticated customers in the world. That's not a bad thing, but if your needs are less, if you're just entering into the data warehousing market … we have just now come out with what we call our data warehouse appliance," he indicates. "It doesn't have all of the scalability features -- like the ability to support mixed workloads or the availability and concurrency features -- of our high-end platforms, but it does meet the data warehousing and the business intelligence needs of customers just entering into the space."
Teradata has long been an advocate of active data warehousing -- or what competitors like to deride as everything-and-the-kitchen-sink data warehousing. It champions an approach in which organizations push to un-silo data and consolidate it in an enterprise-wide data warehouse. This approach is admirable (indeed, desirable) in practice, skeptics concede, but difficult (if not impossible) to achieve in practice.
Even so, Lea maintains, there's a clear sense that the industry is starting to come around to Teradata's active data warehousing model. The proof, he argues, is that customers who buy into Teradata's lower-end appliances frequently end up pursuing -- or taking steps toward -- an active data warehousing strategy.
"It's the value of integrated data. More and more customers want to bring the data together. We have a set of slides that start out with two data marts. In the next slide, those two [data marts] overlap -- and that [overlap] is the number of questions you can answer if you combine the data marts. If you bring in another subject area and another subject area, you can answer more and more questions," he argues.
In an ideal world, Lea insists, everyone would opt for a one-stop data warehouse. "If you had to start off building your analytical environment from scratch, everybody would put it into a central data warehouse, so why do people not put it into a central data warehouse all of the time?" he continues. "Everybody has the vision and is doing it, but hey have to deviate from it from time to time. For political reasons, for compliance, IT has to deviate and do its own thing. There are more customers that are trying to get to the integrated view; there's more of a focus around getting there."
Teradata also tested the waters at PARTNERS last week when it showcased a new SSD-based data warehouse array.
"It's our 'concept car' data warehouse, but it's actually up and running. We went out and bought some very expensive solid state drives. We just wanted to demonstrate that our technology can scale in innovative ways," he explains.
SSD-powered Teradata will make it easier for customers to pursue a "temperature-based" performance model, in which they're able to move frequently accessed data onto a fast (e.g., SSD) tier and less frequently (or only occasionally) accessed information onto a slower tier (e.g., high-capacity 7200 RPM storage).
This capability will be enabled by Teradata 13, the next revision of the flagship Teradata database. That product boasts improved replication capabilities and a new Teradata Virtual Storage facility, among other features.
"Teradata 13 and Teradata Virtual Storage will come out in the first half of next year, and they will actually move data according to temperature, so now we will be able to have multi-tiered storage within a single data warehouse," Lea indicates.
"We'll probably be releasing an SSD-only appliance at some point, but probably not before 2011. It'll be very fast, but it won't be very large. [Teradata virtual storage] will move the hot data to the faster SSDs, and the lesser-used data, the cold data, to the physical drives. It's all done automatically. You don't have to have someone determine what data to place where."