Hi all,
         we want to launch mapreduce job by our java program in client machine, the mapreduce job job jar and lib jar is put on hadoop cluster, currently, we use oozie to launch job and put jar on hdfs . Our mapreduce job is quite simple and sometimes it is just a small job,however, oozie job need to be started with oozie launcher which increase execution time for our mapreduce job , especially for some small jobs, our application require quick response when execute those small jobs, and we don't need any complex workflow functionality to schedule the job, so are there any other way we can do to run a mapreduce job remotely by a java program and jar is pre-installed on hadoop cluster?