Teradata ports Aster analytics to Hadoop Posted on : Aug 31 - 2016

Teradata has at long last decoupled Aster analytics from the underlying database. It reflects the fact that columnar databases alone won't differentiate a product.

With the new version of Aster analytics, Teradata has finally done the inevitable. The new release makes the unique analytic functions of Aster available as software-only for Hadoop and the Amazon Web Services cloud. It decouples Aster analytics from the Aster database; until now, you had to buy the Aster database to get the Aster analytics capabilities.

This comes just over five years after Teradata acquired Aster, a specialized columnar analytic database that had over a hundred analytic functions, some of them patented. Specifically, Aster analytics allows SQL developers to work with what they know -- SQL -- to get access to advanced analytic techniques that they would use to explore data. Examples of Aster analytic functions include SQL-MapReduce computation for complex multi-step problems; SQL-GR for graph analysis that is used in deciphering the interrelationships of people and/or things; and nPath, which provides path analysis for problems such as analyzing and optimizing how customer navigate online sites.

The new release acknowledges a reality in the market: the columnar database technology that was leading edge and unique five years ago is no longer so, and that the real value-add of the Aster platform is its SQL-based exploratory analytics.

Aster analytics was originally developed for columnar database table architectures and used to differentiate Aster Data from then-rivals like Greenplum and Netezza. Columnar tables are better suited than traditional row-based counterparts for analytic queries, where you care about ranges of values, as opposed to ranges of individual records. A technology pioneered by IQ Systems (later Sybase, and now SAP IQ) a decade earlier, a wave of new providers capitalized on columnar at the time with innovations in data compression and in-memory processing that enabled data warehouses to boost their scalability to gigabyte range when Hadoop was breaking barriers with terabytes and petabytes.

While providers like Teradata initially had to buy their way into columnar technology, today columnar is add-on option for most major household brand relational databases. Columnar has also become the de facto standard for cloud-based data warehousing platforms such as AWS Redshift or Snowflake.

In the meantime, Hadoop has become a much more capable platform, and the logical target for utilizing Aster analytics. In the early days of Aster Data, Hadoop was strictly a MapReduce batch analytics platform. Today, Hadoop can run a variety of workloads from batch to interactive and real-time. And the scalability of Hadoop compute and storage makes it a natural target for Aster analytics.

Before this, you could run Aster analytics on Hadoop through Teradata's QueryGrid, but it required the Teradata Aster database platform to initiate and push down query processing. With the new release, Aster analytics are decoupled from the database. While existing Teradata Aster customers won't get unplugged, the trend is clear that the future of Aster is in the analytics library, not the database.

Teradata is making Aster analytics available on two of the three most logical targets: the Hadoop platform, which provides scalability, and AWS, making Aster analytics available from the cloud on demand. As for running Aster analytics directly on the Teradata data warehouse mother ship, that one will wait. Source