have you verified that all the slaves are running tasks? sometimes only a few slaves on a cluster willl pick up a task because of other limitations. Also some algorithms in mahout arent distribnuted. also obviously you will want to make sure that your running the distributed implementations of these algorithms -
On Tue, May 27, 2014 at 8:45 PM, dongdan39 <dongda...@gmail.com> wrote: > Hi, Expert, > I'm confused about the runtime of mahout on e.g Random Forest(the same > with Logistic Regression): no matter how I set the number of slaves from 2, > 8 to 20 in conf/slaves in Hadoop, > the runtime of the program are basically the same. Shouldn't it be faster > when the program runs on more machines? Any hint? > > Regards, Dong > > -- Jay Vyas http://jayunit100.blogspot.com