Database Reference
In-Depth Information
•
mapred.reduce.parallel.copies
: This property defines the number of
parallel transfers done by reduce during the shuffle phase.
•
mapred.submit.replication
: This property defines the replication factor
for the job-related resources that are copied to HDFS at the initiation of a job run.
•
mapred.reduce.tasks
: This property defines the number of reduce tasks
for a single job.
•
mapred.userlog.retain.hours
: This property defines the retention peri-
od of the user logs after the job completion.
•
mapred.child.java.opts
: This property defines the parameters that are
passed to the tasktracker's child processes.
•
mapred.job.reuse.jvm.num.tasks
: The tasktracker creates a JVM for
each task. This property helps to alter this behavior to run more than one task per
JVM.
•
mapred.map.tasks.speculative.execution
: Speculative execution
is the ability of the jobtracker to identify slow running tasks and start another in-
stance of the same task in parallel. The results of the task that finishes first will be
considered and the incomplete task is discarded. This helps in situations when
nodes that are running tasks face some kind of performance problems. If this
property is set to
true
, two instances of the same map task could run in parallel.
•
mapred.reduce.tasks.speculative.execution
: If this property is
set to
true
, multiple instances of the same reduce task can run in parallel.
•
mapred.reduce.slowstart.completed.maps
: This property defines
the percentage value of how much a map task should complete before the redu-
cers are scheduled for their tasks.
