MapR Technologies, Inc. announced at the O’Reilly Strata Conference a new world record for MinuteSort, sorting 15 billion 100-byte records (a total of 1.5 trillion bytes) in 60 seconds using Google Compute Engine and the MapR Distribution for Apache Hadoop.
The benchmark demonstrates how quickly data can be sorted starting and ending on disks. MapR’s benchmark was completed on 2,103 virtual instances in the Google Compute Engine. Each virtual instance consisted of four virtual cores and one virtual disk, for a total of 8,412 virtual cores and 2,103 virtual disks. The new record surpassed the previous record of 1.4TB. The new record is also almost three times the amount of data processed by the previous Hadoop MinuteSort record which was 578 GB.
“The record is significant because it represents a total efficiency improvement executed in a cloud environment,” said Jack Norris, VP of marketing, MapR Technologies. “In an era where information is increasing by tremendous leaps, being able to quickly scale to meet data growth with high performance makes business analytics a reality in situations previously impossible.”
While the previous MinuteSort record was achieved with custom hardware, MapR set the record using commercially available Google Compute Engine, Hadoop MapReduce and the MapR Distribution.
February 26, 2013
Discussion
There are 0 discussion items posted.
Join the Discussion
To join the discussion, become a registered user today!