Netezza earlier this week released a smaller version of its flagship TwinFin data warehouse appliance that the vendor says is ideal for departmental-level deployments.
Analysts warn, however, that the danger of creating data silos still exists.
Netezza Skimmer can store and analyze up to 10 TBs of data, according to Phil Francisco, vice president of marketing at the Marlborough, Mass.-based vendor. The appliance is built on the same architecture as TwinFin, meaning it can be deployed quickly on a departmental level but easily integrated back with a centrally operated TwinFin enterprise data warehouse in a classic hub-and-spoke architecture.
Despite Netezza's claims, however, users still have to architect the system in order to prevent such occurrences, according to Jim Kobielus, an analyst with Cambridge, Mass.-based Forrester Research.
Departmental Skimmer deployments could easily aggravate the departmental silo problem," Kobielus said. "It's still possible for people to shoot themselves in the foot … but this is not necessarily Netezza's fault."
In order to prevent data silos, users must ensure that their departmental data warehouses and data marts are accessing data from common sources with standardized definitions, metadata and data models.
That Skimmer's architecture is based on TwinFin makes the job easier if an organization's enterprise data warehouse is a TwinFin box. But work must still be done behind the scenes, Kobielus said.
"To clean up that mess, you truly need a master data management environment," he said. "It's all about data modeling, having common rules, definitions and hierarchies, then implementing them in all your data marts and data warehouses."
Wayne Eckerson, director of research and services for The Data Warehouse Institute in Renton, Wash., agreed. Without proper design work ahead of time, any departmental data warehouse or data mart deployment risks creating data silos, he said.
"Any company-shared data should reside first in the enterprise warehouse and then be extracted for departmental warehouses," Eckerson said.
Even departmental data warehouse deployments that access and analyze only data specific to particular departments should be built in such a way that they can easily be connected to outside data sources when and if necessary, he said. "You're going to enclose yourself in a silo if you don't design with an eye to the future."
Skimmer deployments at small and medium-sized businesses (SMBs) could be a more practical approach, Forrester's Kobielus said. Skimmer's pricing could make it a viable option for smaller organizations that heretofore couldn't afford enterprise-level analytic databases, even as their prices have been coming down.
Netezza did not release pricing information, but Skimmer will probably check in around $13,000 per terabyte, according to Kobielus. That's around $7,000 per terabyte less than TwinFin and comparable enterprise-level data warehouses from Teradata and Oracle, which can scale to the petabyte level, he said.
But Netezza has company in the SMB data warehouse market, including Teradata, which last year released a series of smaller data warehouse appliances aimed at SMBs and departmental deployments. And Microsoft's soon-to-be-released SQL Server 2008 R2 Parallel Data Warehouse, previously called Project Madison, is also likely to fall somewhere around the $13,000 per terabyte price range.
With Skimmer, Netezza is also targeting ISVs that are looking to integrate analytic capabilities into their software offerings, the vendor's Francisco said. For example, Kalido's Kona, a preconfigured bundle of data integration and analysis software, is built on top of Skimmer.
"And it's not just Kalido," Kobielus said. "Netezza has a lot of partners … that are building applications for various vertical markets. Skimmer is built as a development platform."