使用ycsb测试mongodb

  YCSB（Yahoo! Cloud Serving Benchmark）是雅虎开源的一款通用的、针对各类NoSQL产品的性能测试工具。截止目前为止本人使用YCSB对Mongodb、Redis进行相应的基准测试。本文将介绍测试mongodb完整过程。
1、测试环境准备，所用软件有：
    JDK1.7
    apache-maven-3.1.1
    mongodb-2.6.11
    准备过程：
    

点击(此处)折叠或打开

1. ##配置JAVA环境
   
2. [root@mas ~]# rpm -ivh jdk-7u79-linux-x64.rpm
3. Preparing... ########################################### [100%]
4. 1:jdk ########################################### [100%]
5. Unpacking JAR files...
6. rt.jar...
7. jsse.jar...
8. charsets.jar...
9. tools.jar...
10. localedata.jar...
11. jfxrt.jar...
12. 
    
13. vi  /etc/profileexport JAVA_HOME=/usr/java/jdk1.7.0_79   //添加如下三行、使配置生效、测试
14. export JAVA_HOME=/usr/java/jdk1.7.0_79   
15. export PATH=${JAVA_HOME}/bin:${PATH}
16. export CLASSPATH=.:$JAVA_HOME/lib/dt.jar:$JAVA_HOME/lib/tools.jar
17. [root@mas ~]# which java
    
18. /usr/java/jdk1.7.0_79/bin/java
19. [root@mas ~]# java -version
20. java version "1.7.0_79"
21. Java(TM) SE Runtime Environment (build 1.7.0_79-b15)
22. Java HotSpot(TM) 64-Bit Server VM (build 24.79-b02, mixed mode)
23. 
    
24. ##配置MAVEN
25. cd  /usr/local/
26. tar -zxvf  /root/apache-maven-3.1.1-bin.tar.gz 
27. ln -s  /usr/local/apache-maven-3.1.1   /usr/local/maven
28. vi  /etc/profile.d/maven.sh
29. [root@mas local]# cat  /etc/profile.d/maven.sh   
30. export M2_HOME=/usr/local/maven
31. export PATH=${M2_HOME}/bin:${PATH}
32. 
    
33. [root@mas local]# mvn -version
34. Apache Maven 3.1.1 (0728685237757ffbf44136acec0402957f723d9a; 2013-09-17 23:22:22+0800)
35. Maven home: /usr/local/maven
36. Java version: 1.7.0_79, vendor: Oracle Corporation
37. Java home: /usr/java/jdk1.7.0_79/jre
38. Default locale: en_US, platform encoding: UTF-8
39. OS name: "linux", version: "2.6.18-308.el5", arch: "amd64", family: "unix"
40. 
    
41. 解压已编译好的ycsb,ycsb可以去github下载
42. cd /server/
43. tar -zxvf  /root/ycsb-0.4.0.tar.gz 
44. 
    
45. 解压下载的mongodb数据库软件，并启动mongodb实例
46. [root@mas server]# tar -zxvf /root/mongodb-linux-x86_64-2.6.11.tgz 
    
47. [root@mas server]# mv mongodb-linux-x86_64-2.6.11  mongodb-2.6.11
48. [root@mas server]# ./bin/mongod --dbpath=/server/mongodb-2.6.11/data/ --logpath=/server/mongodb-2.6.11/error.log --logappend &
49. [root@mas server]# netstat -ntlp|grep mongod
50. tcp        0      0 0.0.0.0:27017               0.0.0.0:*                   LISTEN      3331/mongod 
51. 
    
    

2、测试mongodb 2.7.11版本

点击(此处)折叠或打开

1. ##编辑测试模板
2. [root@mas ~]# cd /server/ycsb-0.4.0/
3. vi workloads/workloada  //可参考YCSB提供的模板自行定义，此次测试模板内容如下
4. mongodb.url=mongodb://localhost:27017
5. mongodb.database=ycsb
6. mongodb.writeConcern=normal
   
7. recordcount=1000000
8. operationcount=1000000
9. workload=com.yahoo.ycsb.workloads.CoreWorkload
10. readallfields=true
11. readproportion=0.6    //read操作比例
12. updateproportion=0.3  //update比例
13. scanproportion=0 
14. insertproportion=0.1  //insert比例
15. requestdistribution=zipfian
16. 
    
17. ##
18. ##load数据,建议升级Python到2.7版本，貌似argparse模块被添加Python2.7标准库了，先前版本是没有，得手动安装，否则在load数据这一步会报错，提示缺少该模块
19. [root@mas ycsb-0.4.0]# ./bin/ycsb load mongodb  -s -P  workloads/workloada    //可以-threads参数开启多线程
20. 
    
21. 
    
22. [root@mas ycsb-0.4.0]# ./bin/ycsb run mongodb  -s -threads 16 -P  workloads/workloada  
23. ##经过大量时间测试后，最后输出测试结果：整个测试时间，增删改查平均、最小、最大延迟，也可以计算出其他技术指标，例如QPS,TPS
24. [OVERALL], RunTime(ms), 2218373.0
25. [OVERALL], Throughput(ops/sec), 450.78081999735844
26. [CLEANUP], Operations, 16.0
27. [CLEANUP], AverageLatency(us), 7701.9375
28. [CLEANUP], MinLatency(us), 3.0
29. [CLEANUP], MaxLatency(us), 123199.0
30. [CLEANUP], 95thPercentileLatency(us), 18.0
31. [CLEANUP], 99thPercentileLatency(us), 123199.0
32. [INSERT], Operations, 100240.0
33. [INSERT], AverageLatency(us), 37795.81236033519
34. [INSERT], MinLatency(us), 313.0
35. [INSERT], MaxLatency(us), 7954431.0
36. [INSERT], 95thPercentileLatency(us), 213503.0
37. [INSERT], 99thPercentileLatency(us), 418559.0
38. [INSERT], Return=0, 100240
39. [READ], Operations, 599577.0
40. [READ], AverageLatency(us), 25763.266649654674
41. [READ], MinLatency(us), 145.0
42. [READ], MaxLatency(us), 7454719.0
43. [READ], 95thPercentileLatency(us), 128383.0
44. [READ], 99thPercentileLatency(us), 352255.0
45. [READ], Return=0, 599577
46. [UPDATE], Operations, 300183.0
47. [UPDATE], AverageLatency(us), 53808.972996472156
48. [UPDATE], MinLatency(us), 218.0
49. [UPDATE], MaxLatency(us), 8138751.0
50. [UPDATE], 95thPercentileLatency(us), 282111.0
51. [UPDATE], 99thPercentileLatency(us), 528383.0
52. [UPDATE], Return=0, 300183
    
53.  
54. 
    

3、测试mongodb3.0.7版本，对上面测试mongodb.2.7.11的结果进行对比，验证其性能的提升，本人测试机器差劲，估计没戏

点击(此处)折叠或打开

1. [root@mas server]# ./mongodb-3.0.7/bin/mongod --logappend --logpath=/server/mongodb-3.0.7/error.log --dbpath=/server/mongodb-3.0.7/data &
2. [root@mas server]# ps -ef|grep mongo   
   root      6681  3733  7 12:42 pts/1    00:00:23 ./mongodb-3.0.7/bin/mongod --logappend --logpath=/server/mongodb-3.0.7/error.log --dbpath=/server/mongodb-3.0.7/data
   root      6715  3772  0 12:46 pts/3    00:00:00 python ./bin/ycsb load mongodb -s -threads 16 -P workloads/workloada
   root      6716  6715 83 12:46 pts/3    00:00:25 /usr/java/jdk1.7.0_79/bin/java -cp /server/ycsb-0.4.0/mongodb-binding/conf:/server/ycsb-0.4.0/conf:/server/ycsb-0.4.0/lib/jackson-core-asl-1.9.4.jar:/server/ycsb-0.4.0/lib/jackson-mapper-asl-1.9.4.jar:/server/ycsb-0.4.0/lib/HdrHistogram-2.1.4.jar:/server/ycsb-0.4.0/lib/core-0.4.0.jar:/server/ycsb-0.4.0/mongodb-binding/lib/logback-classic-1.1.2.jar:/server/ycsb-0.4.0/mongodb-binding/lib/mongodb-async-driver-2.0.1.jar:/server/ycsb-0.4.0/mongodb-binding/lib/mongodb-binding-0.4.0.jar:/server/ycsb-0.4.0/mongodb-binding/lib/mongo-java-driver-3.0.3.jar:/server/ycsb-0.4.0/mongodb-binding/lib/logback-core-1.1.2.jar:/server/ycsb-0.4.0/mongodb-binding/lib/slf4j-api-1.6.4.jar com.yahoo.ycsb.Client -db com.yahoo.ycsb.db.MongoDbClient -s -threads 16 -P workloads/workloada -load
   root      6768  3733  0 12:47 pts/1    00:00:00 grep mongo
3. [root@mas ycsb-0.4.0]# ./bin/ycsb run  mongodb  -s -threads 16 -P  workloads/workloada 
4. 测试结果如下：
5. [OVERALL], RunTime(ms), 2487847.0
6. [OVERALL], Throughput(ops/sec), 401.95397868116487
7. [CLEANUP], Operations, 16.0
8. [CLEANUP], AverageLatency(us), 43344.75
   
9. [CLEANUP], MinLatency(us), 2.0
   
10. [CLEANUP], MaxLatency(us), 626687.0
    
11. [CLEANUP], 95thPercentileLatency(us), 67071.0
    
12. [CLEANUP], 99thPercentileLatency(us), 626687.0
    
13. [INSERT], Operations, 100249.0
    
14. [INSERT], AverageLatency(us), 30794.44838352502
    
15. [INSERT], MinLatency(us), 316.0
    
16. [INSERT], MaxLatency(us), 2162687.0
    
17. [INSERT], 95thPercentileLatency(us), 148479.0
    
18. [INSERT], 99thPercentileLatency(us), 348927.0
    
19. [INSERT], Return=0, 100249
    
20. [READ], Operations, 599559.0
    
21. [READ], AverageLatency(us), 35769.579687737154
    
22. [READ], MinLatency(us), 161.0
    
23. [READ], MaxLatency(us), 2160639.0
    
24. [READ], 95thPercentileLatency(us), 196735.0
    
25. [READ], 99thPercentileLatency(us), 421119.0
    
26. [READ], Return=0, 599559
    
27. [UPDATE], Operations, 300192.0
    
28. [UPDATE], AverageLatency(us), 50238.05311600576
    
29. [UPDATE], MinLatency(us), 227.0
    
30. [UPDATE], MaxLatency(us), 2162687.0
    
31. [UPDATE], 95thPercentileLatency(us), 270335.0
    
32. [UPDATE], 99thPercentileLatency(us), 518655.0
    
33. [UPDATE], Return=0, 300192
      
    
    

由于我测试用是N年前被淘汰机器的，CPU还是奔四的，从两次测试的结果来看没多大区别，哪天换台PC Server试试，这里主要记录下测试的整个过程。

参考自：https://github.com/brianfrankcooper/YCSB/tree/master/mongodb