Fujitsu has unveiled a new column-oriented data-processing engine that brings up to a 50-fold increase in data processing speeds across database systems.
The new engine, which runs on a PostgreSQL open-source database, to eliminate the problems that come from trying to reflect the updates or changes to row-oriented data in column-oriented data.
Without being dependent on memory capacity, the engine updates column-oriented data in response to changes in row-oriented data whilst still being able to process column-oriented data at a high speed. It does this by quickly analysing the indexes provided by most database systems and also allows developers determine easily whether a storage method is column-oriented or row-oriented.
Implementation in 2015
The uplift in speed comes from a parallel-processing engine specifically suited to processing column-oriented data that allows analyses to run on a single CPU core to be conducted at four times the speed. For a server equipped with 15 CPU cores this translates to analyses that can be run at least 50 times faster.
Fujitsu is currently aiming for commercial implementation of the new technology during fiscal 2015 as part of its Symfoware Server database product.