Any machine that has access to your namenode and jobtracker is fine. Often people set up one or more machines specifically for launching jobs. Or in a small cluster you can use the NN/JT machine as previously suggested.
Alan. On Feb 3, 2012, at 12:35 PM, Gabriel Rosendorf wrote: > Hi, I'm setting up Hadoop and Pig to read data out of a Cassandra cluster. I > have TaskTrackers installed on each Cass node, and a separate > NameNode/JobTracker. Where does Pig fit into the puzzle? Should it be > installed on the Namenode? Or on one of the Cass nodes with the TaskTracker? > > Thanks, > Gabriel Rosendorf
