Pig provides mean/min/max time per mapper and reducer per job, so you can multiply that out I suppose. Gets a bit tricky with parallel jobs.
On Wed, May 11, 2011 at 6:06 PM, Kris Coward <k...@melon.org> wrote: > > Is there any tool (or even just a good starting point in the logs) for > measuring the amount of cluster used by a pig job (i.e. something that > can be checked after a job was run on a cluster that was already running > other jobs, and provide an indication how much time the job would've > taken if the cluster had otherwise been idle)? > > Thanks, > Kris > > -- > Kris Coward http://unripe.melon.org/ > GPG Fingerprint: 2BF3 957D 310A FEEC 4733 830E 21A4 05C7 1FEB 12B3 >
