hello,
Since I have to compare hadoop performance to other technologies like MPI i need to know how and what to measure. in hadoop the web ui for the jobtracker provides for each successful job information like: Started at: Sat Apr 12 22:34:24 CEST 2008 Finished at: Sat Apr 12 22:35:03 CEST 2008 Finished in: 38sec what exactly does that mean? how are these times determined? when is a job finished? which event 'triggers the clock' to stop and say: 'hey, the job is finished!'. console output (saying 'job complete') takes longer than (in that case) 38 secs. how does the shell script check if a job is finished and what does 'finished' actually mean? are there any special signals around? thank you for any detailed information. -- View this message in context: http://www.nabble.com/time-measuring-tp16656657p16656657.html Sent from the Hadoop core-user mailing list archive at Nabble.com.
