Feature Or Bug RSS

A tumblelog of geeky things.

 Compiled by Ian Sefferman.

Archive

Jan
6th
Sun
permalink

MapReduce Stats

As much as I’m sick of the MapReduce hype, these are impressive stats:

  • 100,000 MapReduce jobs executed per day
  • 20 petabytes of data processed per day
  • 10,000 MapReduce programs written
  • 11,081 machine years used in September, 2007
  • 2,200,000 MapReduce jobs executed in September, 2007 
  • Average of 400 machines per job
  • Average of 395 second completion time per job
(as reported in MapReduce: Simplified Data Processing on Large Clusters, via Greg Linden.
Comments (View)
blog comments powered by Disqus