Hadoop 集群测试报告
1.测试环境
硬件环境
软件环境
o RHEL6.2 2.6.32-220.el6.x86_64
o hadoop-1.0.2
o hive-0.8.1
o hbase-0.92.1
o Oracle 11g_r2(64 位)
o KVM
o MySQL-server-5.5.23-1.el6.x86_64
o jdk-7u3-linux-x64
o sqoop-1.4.1-incubating__hadoop-1.0.0.tar.gz
2.集群部署
3.Hadoop 集群基准测试
3.1 写测试
写 10 个 1000M 的文件
[root@hadoop1 hadoop]# hadoop jar hadoop-test-1.0.2.jar TestDFSIO -write -nrFiles 10 -fileSize
1000
结果:
----- TestDFSIO ----- : write
Date & time: Tue Jul 10 10:52:56 CST 2012
Number of files: 10
Total MBytes processed: 10000
Throughput mb/sec: 19.991963230781227
Average IO rate mb/sec: 20.200298309326172
IO rate std deviation: 2.094306472547166
Test exec time sec: 104.689
The index 1 ≤ i ≤ N denotes the individual map tasks
hadoop6
taskTracker
regionserver
datanode
hadoop5
hadoop4
hadoop3
hadoop2
hadoop22
taskTracker
regionserver
datanode
hadoop21
taskTracker
regionserver
datanode
hadoop31
taskTracker
regionserver
datanode
hadoop32
taskTracker
regionserver
datanode
hadoop41
taskTracker
regionserver
datanode
hadoop42
taskTracker
regionserver
datanode
hadoop51
taskTracker
datanode
regionserver
hadoop52
taskTracker
datanode
regionserver
hadoop1
JobTracker
Hmaster
namenode
secondary namenode
3.2 读测试
读 10 个 1000M 的文件
[root@hadoop1 hadoop]# hadoop jar hadoop-test-1.0.2.jar TestDFSIO -read -nrFiles 10 -fileSize
1000
结果:
----- TestDFSIO ----- : read
Date & time: Tue Jul 10 11:03:14 CST 2012
Number of files: 10
Total MBytes processed: 10000
Throughput mb/sec: 41.544111537630656
Average IO rate mb/sec: 44.81391143798828
IO rate std deviation: 13.201496268890057
Test exec time sec: 67.595
清除测试数据
[root@hadoop1 hadoop]#hadoop jar hadoop-test-1.0.2.jar TestDFSIO -clean
3.3 排序测试
1. 在各节点上分别运行 10 个 map 任务,每个节点产生 10GB 大小的随机二进制数据。
# hadoop jar hadoop-examples-1.0.2.jar randomwriter /benchmarks/random-data
MapReduce(map:90 个,reduce:0 个) Finished in: 14mins, 37sec
2. 排序
#hadoop jar hadoop-examples-1.0.2.jar sort /benchmarks/random-data
/benchmarks/sorted-data
MapReduce(map:720 个,reduce:48 个) Finished in: 1hrs, 34mins, 15sec
3. 检查是否已经排好序
#hadoop jar hadoop-test-1.0.2.jar testmapredsort -sortInput /benchmarks/random-data
-sortOutput /benchmarks/sorted-data
MapReduce(map:138 个,reduce:1 个)Finished in: 14mins, 25sec
结果提示:SUCCESS! Validated the MapReduce framework's 'sort' successfully. 说明都已
排序
3.4MapReduce 连续性测试
每个 JOB 处理 10000 行文本文件,6 个 map,2 个 reduce,只对数据进行收集,总共跑 10
次。
#hadoop jar hadoop-test-1.0.2.jar mrbench -numRuns 10 -maps 6 -reduces 2 -inputLines
10000