Aiming to help organizations orchestrate analytical processing across myriad systems without moving data, Teradata unveiled Teradata QueryGrid at its Teradata Universe show in Prague today.
QueryGrid is a software-based service embedded in Teradata Database 15, also announced today, that's designed to allow Teradata Database to intelligently use the functionality and data of multiple heterogeneous processing engines, including Hadoop (via remote, push-down processing), other Teradata Databases and Teradata Aster Databases, as well as RDBMS databases.
"Attempts at federation have been unsuccessful for many reasons," says Scott Gnau, president of Teradata Labs. "To deliver value from big data, customers should create an architecture that allows the orchestration of analytic processes across parallel databases rather than federated servers. After the user selects an analytic engine and a file system, Teradata software seamlessly orchestrates analytic processing across systems with a single SQL query, without moving the data. In addition, Teradata allows for multiple file systems and engines in the same workload."
Queries can be initiated from a Teradata Database to access, filter and return subsets of data from Hadoop, Aster and other database environments to the Teradata Database for additional processing.
For instance, Gnau says, you might have a fraud analysis that's running in your data warehouse. To supercharge that process, you want to extend the analysis to seek out anomalies in your Web transactions — Web log data that probably resides in your Hadoop cluster. QueryGrid can spawn an algorithm that runs in your Hadoop cluster and then delivers the answer to your data warehouse, which can then incorporate the results into the final analysis.
"Teradata pioneered integration with Hadoop and HCatalog with Aster SQL-H to empower customers to run advanced analytics directly on vast amounts of data stored in Hadoop," says Ari Zilka, CTO of Hortonworks, whose engineering team works closely with Teradata engineers.
"Now they are taking it to the next level with pushdown processing in Hadoop, leveraging the Hive performance improvements from Hortonworks' Stinger initiative, delivering results at unprecedented speed and scale," Zilka says.
Gnau says QueryGrid capabilities will be available in the third quarter.