I got an update from the folks at Teradata on the Teradata Aster platform recently. I have been briefed on various partnerships around Teradata before – they are very active working with analytics providers on making sure that in-database analytic engines work on Teradata – but had not been briefed specifically on Teradata Aster.
Aster was acquired by Teradata back in 2011 and has recently announced their Aster Discovery Platform 5.10. Teradata Aster’s core is an implementation of MapReduce inside their database that is made available through a traditional SQL interface. This allows for multi-pass, procedural elements to be combined with SQL in accessing a database that stores both structured and unstructured data. Aster has a wide range of SQL-MapReduce functions pre-built and 5.10 adds a new category of in-database, visual SQL-MapReduce ® function for advanced visualizations. The visual SQL-MapReduce ® function include – Flow Visualizer, Affinity Visualizer and Hierarchy Visualizer.
Teradata Aster is very focused on making it easier for users to build analytic applications that combine acquiring, preparing, analyzing and visualizing data.. Aster’s pre-built library of SQL-MapReduce functions will enable customers to go from data all the way to visualizations quickly through a single SQL-MapReduce query. These functions are designed to support the kind of analyses that are common in the Teradata Aster user community:
- Flow or path analysis used, for instance, in identifying paths customers take before they churn and showing the concentration of results into the different paths
- Affinity analysis to see what sells or is used with what, shown using affinity chords that visualize the strength of relationships between things
- Hierarchies such as rates of shopping cart abandonment by product where products are part of a deep, complex hierarchy.
Besides these visualizations, the new release also integrates with Attensity. Attensity is a well established text analytics company. The integration offers all the Attensity functions, executing them in-database to support text analysis of data stored in Teradata Aster. Queries can use these Attensity functions directly. Attensity and Teradata Aster have some joint customers already and Teradata really liked Attensity’s support for non-English languages.
5.10 also integrates with the Zementis ADAPA engine to allow the in-database scoring of predictive analytic models described using PMML. The interface with Zementis can be used in batch to score data records or it can be used as a function in a query to support real-time scoring.
5.10 also begins Teradata Aster’s support for R. Currently this allows for some R functions to be executed in database and R scripts can be executed against Teradata Aster data in a parallel fashion via Aster’s SQL-MapReduce interface In the future Aster will have a more seamless integration with R client in supporting in-database R modeling and Teradata Aster is considering support for additional R distributions besides the base open source R.
A key element in the Teradata Aster value proposition remains the ability to manage analysis of many different kinds of data from logs to customer database with many kinds of analysis/access from SQL to MapReduce, time series, geo spatial etc. For example customer interaction data from many channels (branch, call center, ATM, web) needs to be analyzed to find churn patterns. There are many paths to churn that can be analyzed and visualized. Teradata Aster allows customers to do this by refining a single SQL MapReduce query. One customer, having found the paths to churn this way, were able to improve their models for churn by using Teradata Aster queries that find customers on these paths and combined this information in Teradata EDW with other data for churn modeling.
More information on Teradata Aster is available here.