Oracle’s Software in Silicon Technology: Data Analytics Accelerators (DAX)



this is a demonstration of Oracle's software and silicon data analytics accelerators commonly called Dax high-performance big data analytics are now critical to surviving in today's real-time business environment this overwhelming growth of data means that not only must you be able to analyze the information that truly counts but you also must involve sophisticated search and machine learning algorithms over large data sets Oracle is rethinking this challenge with their innovative approach to processor design Oracle's revolutionary software and silicon technology hardwires key software processes directly onto the processor to enable high-speed analytics and secure data without adding latency Dax R of revolutionary software and silicon innovation which dedicate coprocessors on to each spark microchip to process specific types of analytic operations the dax coprocessors located on either side of the SPARC processor process large data sets directly from memory leaving the CPU cores to focus on other tasks this approach reduces data transfers offload CPU tasks and reduces level three cache pollution as a result operations that are common in big data analysis machine learning algorithms and in Oracle database in-memory large-scale queries all see significant performance improvements with Dax each spark processor has eight Dax coprocessors with four data pipelines each this configuration enables the chip to process 32 data streams that are completely independent from the work performed by the processor core to see how offloading SQL queries to the Dax coprocessors produces such dramatic performance benefits let's take a look at the way an in-memory analytically runs in this example an SQL query runs only on the CPU core the core reads the data which is already compressed in memory decompresses it writes it back to memory again from memory and then performs a software scan and writes back the results this process can inhibit performance now let's take a look at the way an in-memory analytics runs when Dax is in use in this example Dax decompresses data on the fly this real-time decompression eliminates the iterative steps of traditional SQL query processing which simplifies the operation and eliminates clutter in the processor cache the net result is that queries are processed as much as ten times faster than queries run in the traditional manner note that when you enable the Oracle database in-memory feature in Oracle Database 12c analytics and mixed OLTP workloads automatically benefit from improved performance and real-time analytics Dax can also be used through application programming interfaces called api's to accelerate a variety of analytic workloads Apache spark is a popular open source clustering computing framework from the Apache Software Foundation by allowing user programs to load data into a clusters memory then query it repeatedly SPARC is well-suited to machine learning algorithms and iterative analytic processes it is this performance gain that has made Apache spark one of the most popular methods for processing large data sets through open api's Oracle engineers have demonstrated that software and silicon Dax can be used to accelerate Apache spark performance in this project the engineers use spark m7 Dax with Apache spark to take 1 billion rows of data in memory and filter it into a 3d cube to enable interactive data analytics performance testing was performed both with and without Dax integration with Apache spark with Dax enabled the team was able to achieve a 6 times improvement in cube generation performance the analytics queries were handled by Dax freeing the processor cores up for other workloads minimal changes were required to Apache spark to take advantage of the Dax capabilities and packages which are built on top of Apache spark will also benefit from this acceleration with no changes require the phenomenal growth of big data analytics is driving the need for innovation in processor technologies the sparked m7 processor with Dax coprocessors offers unprecedented analytics efficiency for both the Oracle database in memory along with a variety of other applications through open Dax ap is shown here are just a few additional application examples for Dax through Oracle's software and silicon developer program developers can now apply the Dax technology to a broad spectrum of challenges in the analytic space go to Swiss dev dot oracle.com forward slash Dax Dax and caps and leverage Oracle software and Silicon cloud to gain access to this technology along with numerous free developer resources you

One Comment

  1. keepit1oo said:

    Nicely done. Check out my channel if you get the chance, stay blessed!

    June 27, 2019
    Reply

Leave a Reply

Your email address will not be published. Required fields are marked *