September 15, 2010 - Data management and analytics provider Aster Data today announced the release of nCluster 4.6, a hybrid column/row data store that leverages MapReduce and a massively parallel processing database management system.
What's new in 4.6 is the columnar data store, in concert with Aster's existing row store in a heterogeneous platform that makes use of Aster's existing SQL MapReduce technology and MPP. Columnar databases and embedded applications have found renewed attraction in certain use cases involving very large sets of data where processing efficiency carries a premium.
The hybrid function of 4.6 includes a new data modeling "Express" tool, which offers recommendations on what data is best stored in row or column format for optimal performance against workloads and unified access to the resulting data store
Aster data EVP Sharmila Mulligan claims the configuration is first to market and allows richer analytic processing against very large data sets.
"There are many use cases where you would not want to put just your data in a row or column format only because of they type and speed of the analytics you would want some of your data in each," says Mulligan. "The big news is, yes, we're adding a column store but what we're bringing with that is the ability to do even richer analytics through the MapReduce processing technology."
Increasing volumes of data reflect the need for both ad hoc and structured queries. In general, row stores work best for ad hoc or interactive queries while column stores perform better in repetitive, reporting-style queries where the user already knows what data will be retrieved.
Tasso Argyros, CTO and co-founder of Aster Data says having both storage approaches in the same system makes it easy to move data from one system to the other and advise how to best load and store the data.
"Some queries may use both storage systems," Argyros says. "But the decision of where the tools put the data is completely transparent to the analyst who accesses the data. Row or column, the queries run exactly the same way, your BI tools operate the same way, your MapReduce analytical apps run the same way."
Mulligan says nCluster 4.6 works with relational and non-relational data queries; the release is a platform upgrade and an upgrade path for existing customers.