[ https://issues.apache.org/jira/browse/MAPREDUCE-1902?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12891532#action_12891532 ]
Vinod K V commented on MAPREDUCE-1902: -------------------------------------- Both are equally efficient I think, unless you bring in sharing of job jars across jobs also. It'd definitely help code reuse. I checked trunk and realized that only a minor difference exists between the present way and the dist-cache way. We also un-jar the job.jar so that classes inside sub-directories (according to a job-configurable pattern), for e.g., lib/, classes/, are also made to be available on class-path. Accommodating it should be straight forward. > job jar file is not distributed via DistributedCache > ---------------------------------------------------- > > Key: MAPREDUCE-1902 > URL: https://issues.apache.org/jira/browse/MAPREDUCE-1902 > Project: Hadoop Map/Reduce > Issue Type: Improvement > Reporter: Joydeep Sen Sarma > > The main jar file for an job is not distributed via the distributed cache. It > would be more efficient if that were the case. > It would also allow us to comprehensively tackle the inefficiencies in > distribution of jar files and such (see MAPREDUCE-1901). -- This message is automatically generated by JIRA. - You can reply to this email to add a comment to the issue online.