Teradata Mixes, Matches Database Rows, Columns

Seeking to improve access to data stored in columnar tables and, with it, speed up query performance, big data and marketing applications specialist Teradata Corp. rolled out a hybrid row and column database that it said would allow users to mix and match row and column technologies.

San Diego-based Teradata said May 7 its enhanced hybrid row and column in-database technology could help reduce the amount of data read in wide tables while “preserving pinpoint data access” for real-time queries. The upgrade allows Teradata Database users to streamline single-row data retrieval for faster queries, thereby reducing unnecessary row scanning.

Teradata said the enhancements support mission-critical applications that depend on advanced workload management, high availability, rich SQL along with in-database analytics.

Meanwhile, an “update-in-place” capability allows data to be updated within a row. That functionality eliminates the need to copy updated information at the bottom of the table as required in most columnar databases.

Along with improving query performance, Teradata said it also boosts storage efficiency. In a blog post, the company said “isolating the storage for each column makes data temperature measurements from Teradata Virtual Storage more targeted by applying them to the data in a column rather than a mix of data from all columns in a table.” Hence, “data in the most-used columns naturally becomes hot and migrates to the fastest available storage.”

The greatly compressed columnar data “increases the amount of raw user data that fits into the fast storage,” the company added.

In one use case example, Teradata cited telecommunications providers who capture data on each network call in very wide tables called Call Detail Records. The hybrid row and column capabilities of the Teradata database could be used to support analytical queries on different columns of these very wide tables covering subjects like network quality or regional traffic analyses. At the same time, Teradata added, the hybrid format would also allow direct access to a single row of data used in a call center query during customer support.

The in-database technology responds to massive datasets, growing user populations inside enterprises and performance-driven service level agreements. “Organizations everywhere are under extreme pressure to deliver analyses faster and to more people than ever before,” Teradata noted. That has fueled requirements for faster data warehouse performance to speed business decisions and new business applications while improving system utilization.

In the Teradata approach, only data in the columns required for a query are pulled into memory for processing. That, the company claims, vastly reduces the time-constraining I/O of a row-based approach in which data from all the columns is read.

The company said the hybrid row and column feature would be available on Teradata Database 15.10 by the end of the second quarter of 2015.

Recent items:

Teradata Makes Data Warehouse More Hadoop-ish

Teradata Moves Virtual Data Warehouse Forward With MongoDB