hadoop jar /usr/hdp/2.4.0.0-169/hadoop-mapreduce/hadoop-mapreduce-client-jobclient-2.7.1.2.4.0.0-169.jar TestDFSIO -write -nrFiles 100 -fileSize 100
100个文件，每个100M
----- TestDFSIO ----- : write
Date & time: Fri Nov 17 19:39:04 HKT 2017
Number of files: 100 文件数量
Total MBytes processed: 10000.0 总写入数据量
Throughput mb/sec: 75.77709409999545 总共需要写入的数据量/（每个map任务实际写入数据的执行时间之和（这个时间会远小于Test exec time sec））＝＝》10000/(map1写时间 map2写时间 ...)
Average IO rate mb/sec: 82.62942504882812 每个map需要写入的数据量/每个map任务实际写入数据的执行时间）之和/任务数＝＝》(文件大小/map1写时间＋文件大小/map2写时间 ...)/1000，所以这个值跟上面一个值总是存在差异。
IO rate std deviation: 17.816849548077943 上一个值的标准差
Test exec time sec: 450.367整个job的执行时间
Yarn memory 90%
Throughput mb/sec衡量每个map task的平均吞吐量，
Average IO rate mb/sec 衡量每个文件的平均IO速度。
阿里云：
----- TestDFSIO ----- : write
Date & time: Mon Nov 27 15:58:26 CST 2017
Number of files: 100
Total MBytes processed: 10000.0
Throughput mb/sec: 12.550185052478598
Average IO rate mb/sec: 21.50667953491211
IO rate std deviation: 18.03019124739804
Test exec time sec: 140.708
----- TestDFSIO ----- : read
Date & time: Mon Nov 27 16:04:11 CST 2017
Number of files: 100
Total MBytes processed: 10000.0
Throughput mb/sec: 70.17593105916532
Average IO rate mb/sec: 169.13583374023438
IO rate std deviation: 135.4413193785257
Test exec time sec: 86.042
TestDFSIO -clean 清理
读测试：
hadoop jar /usr/hdp/2.4.0.0-169/hadoop-mapreduce/hadoop-mapreduce-client-jobclient-2.7.1.2.4.0.0-169.jar TestDFSIO -read -nrFiles 100 -fileSize 100
read
Date & time: Sat Nov 18 12:08:07 CST 2017
Number of files: 100
Total MBytes processed: 10000.0
Throughput mb/sec: 139.80734547793142
Average IO rate mb/sec: 144.37628173828125
IO rate std deviation: 23.001677374779344
Test exec time sec: 360.126
总结：可以看出一些常见标准
