我提交了一份9000+ mapper和4096 reducer的工作
每个映射器的输入都是一个hdfs文件，大约10GB（5000万行）。
通常，一个映射器运行25分钟。
然而，总有大约1~10个映射器运行成本为1小时
这些地图制作者＆＃39;清理（）发生在mapper的设置（）后20分钟
合并成本为40分钟
如何加快合并？
一些日志：
2018-05-19 12:03:14,791 INFO [main] org.apache.hadoop.mapred.Merger: Merging 8 sorted segments
2018-05-19 12:03:14,792 INFO [main] org.apache.hadoop.mapred.Merger: Down to the last merge-pass, with 8 segments left of total size: 1282276 bytes
2018-05-19 12:03:14,805 INFO [main] org.apache.hadoop.mapred.Merger: Merging 8 sorted segments
2018-05-19 12:03:14,810 INFO [main] org.apache.hadoop.mapred.Merger: Down to the last merge-pass, with 8 segments left of total size: 1181792 bytes
2018-05-19 12:03:14,845 INFO [main] org.apache.hadoop.mapred.Merger: Merging 8 sorted segments
2018-05-19 12:03:14,904 INFO [main] org.apache.hadoop.mapred.Merger: Down to the last merge-pass, with 8 segments left of total size: 1028459 bytes
2018-05-19 12:03:14,912 INFO [main] org.apache.hadoop.mapred.Merger: Merging 8 sorted segments
2018-05-19 12:03:14,913 INFO [main] org.apache.hadoop.mapred.Merger: Down to the last merge-pass, with 8 segments left of total size: 988898 bytes