SGI & Apache Hadoop : Terasort Benchmark – new world record
The Cloudera distribution that includes Apache Hadoop was found to be
81% faster than Oracle Sun X2270 cluster
Fremont, Calif. – October 17, 2011 – SGI, the leader in technical computing, today announced that it has set a new world record on the Terasort Benchmark for processing and analyzing data using Apache Hadoop clusters running on a Cloudera Distribution (CDH). The company, which recently joined the Cloudera Connect Partner Program, also separately announced that it has formed a distribution relationship with Cloudera that will allow it to build, sell and deploy commercial Hadoop-based solutions.
Results achieved in September 2011 show that 20 SGI Hadoop Cluster nodes composed of SGI Rackable C2005-TY6 half-depth servers with Intel ® Xeon E5630 series processors, 48GB of memory, and 4x 1TB SATA HDDs running on the Cloudera CDH3 took only 130 seconds to complete a Terasort with a working size of 100 GB. Terasort helps derive the sort time for 1TB or any other amount of data in a Hadoop cluster, and is a benchmark that combines the HDFS and MapReduce test layers of a Hadoop cluster. In this case, Terasort scales super-linearly on a Rackable SGI C2005-TY6 cluster with Cloudera’s Apache Hadoop deployment (CDH3u0), and has been shown to be 81% faster than a similarly sized Sun Oracle X2270 cluster.
Source : http://www.sgi.com/company_info/newsroom/press_releases/2011/october/hadoop.html
