In late April at San Francisco, Jim Goodnight, CEO of SAS, welcomed attendees to the 38th annual SAS Global Forum by noting that when the conference was first held in San Francisco in 1982, everything was big. Cars were big, hair was big, and so were computers and cell phones. Goodnight celebrated changes in our digital world by noting that many of us now carry the computing power of a 1982 mainframe in our pockets. But while computers have grown smaller, data has gotten much, much bigger and thankfully so has the power of Apache Hadoop and big data analytics solutions. By market share, SAS is the leading business analytics provider in the world, and the SAS annual conference was a great opportunity to take stock of how much and how fast the big data landscape has changed in the last year.
Of course, for Intel the past year's big data highlight was the release of the Intel Distribution for Apache Hadoop* software, and the SAS conference provided a showcase for Intel;s collaborative approach to the Hadoop ecosystem working with other industry leaders, such as SAS, to build an open infrastructure around Hadoop. Goodnight offered an overview of SAS solutions that provide a comprehensive analytic framework for managing large amounts of data in Hadoop. SAS/ACCESS* offers seamless and transparent data access to Hadoop, while SAS Visual Analytics* provides a rich visual and interactive Hadoop experience. SAS High-Performance Analytics Server* works with Hadoop to provide statistical modeling on massive data sets and uncover predictive insights and trends in big data.
These SAS big data analytics solutions all have the potential to integrate with the Intel Distribution for Apache Hadoop, furthering a collaborative environment around Hadoop that will help to ensure that the best data management and analysis solutions are optimized to handle the challenges and reap the rewards of big data. SAS is currently working with Intel toward validation and optimization of its big data analytics solutions with the Intel Distribution.
At our booth, we also demonstrated the new Intel Cache Acceleration Software (Intel CAS) that combines with Intel Solid State Drives (SSDs) to reduce I/O latency and drive targeted application performance for faster big data analytics. Intel CAS installs on the server and uses fast SSD/Flash media and intelligent caching to automatically determine the appropriate cache level for active data and optimize the use of system memory. We bench marked the performance benefits of four different storage configurations (from 100 percent HDD [15K RPM] to 100 percent SSD [37K RPM equivalent], with two hybrid configurations) by running a mixed workload of SAS analytics. Adding Intel CAS acceleration and SSDs to the input partition brought a 30 percent performance gain over a pure, baseline HDD configuration. Adding further SSDs to the SAS Work compute partition boosts performance 95 percent over the HDD baseline. For highest performance, a pure SSD/SSD configuration delivers a 300 percent improvement over the all HDD baseline.
Intel also participated on panel discussions that approached Hadoop from two different directions: Discovering the business value in big data, and how IT orgs can use Hadoop to improve management of big data. Pauline Nist (@panist), general manager of Intel;s Enterprise Software Partner Strategy, joined Cloudera CEO (and soon to be CSO) Mike Olson (@mikeolson) and Paul Kent (@hornpolish), vice president of Big Data at SAS, in leading the discussions. Bringing together executives from different sides of the big data equation made for a mind-opening afternoon of insights, from high-level strategies to hands-on wisdom.
Follow the growing list of data management & #Bigdata @TimIntel followers.