hadoop perform optimization tasks

Original link: http://www.cnblogs.com/mjorcen/p/3965311.html

 Perform optimization tasks

1. speculative execution: 

  If jobtracker found to have delayed the mission, the same will then start a backup job, and then executing the first of which will go to kill off the other, and therefore will often see on the monitor page job done properly executed will be kill off task.

 

2. speculative execution enabled by default, but if the code is the problem, not solve the problem, and will make the cluster more slowly, by setting mapred.map.tasks.speculative.execution and mapred in mapred-site.xml configuration file. reduce.tasks.speculative.execution can be turned on or off speculative execution tasks as map or reduce tasks,

 

3. Reuse jvm.

  The new start time omitted jvm consumed. Mapred.job.reuse.jvm.num.tasks disposed on a single set jvm maximum number (1,> task mapred-site.xml profile 1 or -1 indicates no limit);

 

4. Ignore mode

  After reading task data failed twice, position data will tell jobtracker, and then start the task again and remember to skip when encountered bad data (disabled by default, opened with SkipBadRecord method)

 

Reproduced in: https: //www.cnblogs.com/mjorcen/p/3965311.html

Guess you like

Origin blog.csdn.net/weixin_30648587/article/details/94784729