When Oracle announced its first Sparc M7 servers at Oracle Open World 2015, it mentioned that third-party developers would be granted access to the 'SQL in Silicon' features.
DAX (the new term for this feature, short for 'data analytics accelerator') provides each of the M7's 32 cores with a coprocessor to perform scan, extract, select and translate SQL functions. These coprocessors operate at memory bus speed (up to 160GBps), and are used by Oracle Database 12c.
This speed is achieved in part by the combination of DAX and hardware support for memory decompression.
Vice president of product management for systems Marshall Choy told iTWire that Oracle's engineers had developed code allowing Apache Spark to take advantage of DAX, and that it would be submitted to Apache.
The performance improvement turns what have previously been batch operations into practically interactive activities, he said. DAX allows a database of one billion rows to be analysed interactively.
"In reality, not that much work" is required to DAX-enable a program, said Choy. Adding DAX support to Spark took a week and "the legal agreements took longer," he added
Furthermore, Brown University has been carrying out research into the performance and scalability of DAX in conjunction with Oracle. Computer science department chairman Ugur Centimenel said "We are currently working on characterising the performance of DAX across a suite of modern in-memory data layout schemes.
"After completing this study, we will work on the optimal use of DAX in accelerating interactive data exploration and visualisation with the Tupleware main-memory database system and S-Store real-time stream processing system.
"Through these studies, we will quantify the performance and scalability of M7 and DAX on real workloads involving sophisticated search and machine learning over large data sets."
Choy predicted that "a year from now we'll be looking at use cases that we didn't imagine."
M7 systems "have done exceedingly well" in the market and the open API for DAX will help continue this growth, in part by extending SPARC systems into non-traditional markets, he said.
Interested parties can join Oracle's software in silicon developer program here.