oh..you don't need run several reduce tasks on a small dataset. multiple reduce tasks does not mean a good performance, caus there would be network consumption, and so on. PFP code does not contain such a line to set reduce task number, you can specify it yourself by modifying the code or xml file.
Regards, Q 2012/4/30 培宇 <[email protected]> > the size of dataset only takes dozens MB. > But on my own understanding of PFP, > the more reduce tasks may have better performance, isn't it? > > By the way, I don't find "setNumReduceTasks()" in the source code of Mahout > How does it work with multiple reduce tasks? > > > > > 2012/4/30 戴清灏 <[email protected]> > > > Then how big your input data size is? > > For a rather small dataset, one reduce task is enough to process. > > > > Regards, > > Q > > > > > > > > 2012/4/30 培宇 <[email protected]> > > > > > I set mapred.tasktracker.reduce.tasks.maximum 1 in conf for each node > > > > > > but I have 4 nodes for running Hadoop > > > > > > Should I install mahout for each node or only master node? > > > > > > Thanks for your help > > > > > > > > > 2012/4/30 戴清灏 <[email protected]> > > > > > > > Sorry for having made you confused. > > > > I mean, if you have explicitly specify the reduce task number in your > > > > hadoop conf/mapred-site.xml or some where else, > > > > PFP would only execute one reduce task. > > > > Your parameter groups 10 would only make PFP call reduce method 10 > > times. > > > > Actually reduce method had been called 10 times with a single reduce > > > task. > > > > > > > > > > > > Regards, > > > > Q > > > > > > > > > > > > > > > > 2012/4/30 培宇 <[email protected]> > > > > > > > > > Hello > > > > > I mean reduce tasks. > > > > > I set the parameter -g 10, > > > > > but there is still one reduce task in ParallelFPGrowth. > > > > > > > > > > How do I set parameter -g to change the number of reduce tasks? > > > > > > > > > > Thanks for your reply > > > > > > > > > > > > > > > > > > > > -- > > > > > Best regards, > > > > > 培宇 Omi > > > > > > > > > > > > > > > > > > > > > -- > > > Best regards, > > > 培宇 Omi > > > > > > > > > -- > Best regards, > 培宇 Omi >
