Rather than thinking of various database technologies that have recently gained prominence as alternatives to each other, Teradata is making a strong case for their unification.
Teradata announced that it is extending its Unified Data Architecture by allowing the same SQL programming framework it uses for its relational database and the Aster Data massively parallel database platform to Hadoop, while the same time unveiling a new Fabric approach that serves to logically unify the management of disparate data management systems.
According to Steve Wooledge, vice president of marketing for Teradata Unified Data Architecture, the company is adding an entry-level Teradata Data Mart Appliance 670 that scales up to 12TB, and a high-end Teradata Active Enterprise Data Warehouse 6700 system that scales all the way up to 61 Petabytes using up to 2,000 discrete units of processing running in parallel.
The systems can be configured to run the Teradata SQL database, the Aster Data massively parallel database or the Hadoop distribution from Hortonworks, which can then all be logically managed across a unified Teradata fabric running version 5.0 of its BYNET software from Teradata on top of Infiniband technology developed by Mellanox Technologies.
In addition to the new platforms and SQL support for Hadoop, Teradata is also unveiling Teradata Studio with smart loader for Hadoop. According to Wooledge, this new offering not only makes it easier to browse Hadoop tables, it provides a visual framework for moving data between Hadoop and Teradata systems.
While lots of vendors these days are talking about the unification of Hadoop and traditional data warehouse technologies, Wooledge says that IT organizations should carefully note how many of those approaches wind up forking Hadoop in a way that adds proprietary extensions that eliminate one of the primary benefits of having an open source code base.
In the meantime, it’s clear that the future of the data warehouse is going to be federated well beyond traditional SQL databases. Wooledge says SQL databases will continue to run structured data that most organizations need to rapidly query, while Hadoop will be used to discover relationships between large amounts of unstructured data.
The only real challenge going forward is to figure out how to actually manage all those data sources under a common framework the average database administer (DBA) can effectively master.