Spark should be better than MapReduce (if only it worked) (blog.explainmydata.com)