[MAPREDUCE-5484] YarnChild unnecessarily loads job conf twice

• Type: Improvement
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 2.1.0-beta
• Fix Version/s: 2.3.0
• Component/s: task
• Labels: perfomance
• Hadoop Flags: Reviewed
• Assignee: Sandy Ryza
• Reporter: Sandy Ryza
• Created: 28/Aug/13 08:41
• Updated: 24/Feb/14 20:57
• Resolved: 29/Aug/13 23:54

Description

In MR task processes, a JobConf is instantiated with the same job.xml twice, once at the beginning of main() and once in configureTask. IIUC, the second instantiation is not necessary. These take time reading from disk and parsing XML.

Removing the second instantiation shaved a second off the average map task time in a 1,000-map sleep job.