sysbench性能测试

sysbench的官网地址是：http：//sysbench.sourceforge.net

Step 1： 下载安装： sysbench-0.5.tar.gz 解压后展开目录如下

[root@localhost mnt]# cd  sysbench-0.5
[root@localhost sysbench-0.5]# ls
aclocal.m4  autom4te.cache  config      config.status  configure.ac  doc         libtool  Makefile     Makefile.in  mkinstalldirs  README-WIN.txt  TODO
autogen.sh  ChangeLog       config.log  configure      COPYING       install-sh  m4       Makefile.am  missing      README         sysbench
[root@localhost sysbench-0.5]#

查看README文件可以看到安装方法：

    ./autogen.sh
    ./configure
    make
    make install
 

Sysbench --help可以看到具有以下功能

Compiled-in tests:
  fileio - File I/O test  IO性能测试
  cpu - CPU performance test CPU性能测试
  memory - Memory functions speed test 内存性能测试
  threads - Threads subsystem performance test 线程子系统性能测试
  mutex - Mutex performance test  互斥测试

Step2  Sysbench CPU性能测试：

Sysbench 测试CPU是通过计算素数的加法运算，它会自动生成素数，并进行相加；一般可以指定最大值，默认10000；

我们用命令：sysbench --test=cpu --cpu-max-prime=20000 run

[root@localhost sysbench-0.5]# sysbench --test=cpu --cpu-max-prime=20000 run
sysbench 0.5:  multi-threaded system evaluation benchmark
--版本
Running the test with following options:
Number of threads: 1
--使用一个线程
Random number generator seed is 0 and will be ignored
随机数生成素数0将会被忽略

Primer numbers limit: 20000
最大素数上线20000
Threads started!


Test execution summary:
    total time:                          29.5438s  --总共花费时间
    total number of events:              10000     --一共计算10000次
    total time taken by event execution: 29.5402s  --事件（一个素数计算做一个事件）执行时间
    per-request statistics:
         min:                                  2.48ms  --最小计算花费时间
         avg:                                  2.95ms  --平均
         max:                                 56.82ms  --最大
         approx.  95 percentile:               3.85ms  --95%的计算花费时间

Threads fairness:
    events (avg/stddev):           10000.0000/0.00   计算事件总数
    execution time (avg/stddev):   29.5402/0.00      计算花费时间总数

可以看到压测的时候CPU被sysbench占满了；

Step3: IO测试：

sysbench --test=fileio help

fileio options:
  --file-num=N                  number of files to create [128] 创建文件数量，默认128
  --file-block-size=N           block size to use in all IO operations [16384] 块儿大小，默认16384，即Innodb的存储引擎Page大小；如果要测试Innodb就用默认；
  --file-total-size=SIZE        total size of files to create [2G] 生成测试文件的总大小
  --file-test-mode=STRING       test mode {seqwr（顺序写）、seqrewr（顺序读写）、seqrd（顺序读）、rndrd（随机读）、rndwr（随机写）和rndrw（随机读写）}
  --file-io-mode=STRING          文件IO模式，同步/异步/快速映射/慢映射{sync,async,fastmmap,slowmmap} [sync]。
  --file-extra-flags=STRING     additional flags to use on opening files {sync,dsync,direct} [] 打开文件的额外选项
  --file-fsync-freq=N           do fsync() after this number of requests (0 - don't use fsync()) [100]执行写入fsync()的频率，fsync()用于写入数据导磁盘 默认100次请求就写入一次导磁盘
  --file-fsync-all=[on|off]     do fsync() after each write operation [off]  每次写操作都同步到磁盘
  --file-fsync-end=[on|off]     do fsync() at the end of test [on] 测试结束同步至磁盘
  --file-fsync-mode=STRING      which method to use for synchronization {fsync, fdatasync} [fsync]使用哪种方法写入磁盘
  --file-merged-requests=N      merge at most this number of IO requests if possible (0 - don't merge) [0] 合并改数量的IO请求
  --file-rw-ratio=N             reads/writes ratio for combined test [1.5]。 写/读组合比测试，默认1:1.5

sysbench的fileio测试需要经过prepare、run和clean三个阶段。

1. prepare是准备阶段，生产我们需要的测试文件
2. run是实际测试阶段
3. cleanup是清理测试产生的文

sysbench --test=fileio --file-num=16 --file-total-size=2G prepare  生成16个测试文件，总大小2G

此时会生成测试文件，因为总大小2G，数量16个，所以每个文件约129M；

 sysbench --test=fileio --file-num=16 --file-total-size=2G  --file-test-mode=rndrw run 采用随机读写的方式测试：

[root@localhost mnt]# sysbench --test=fileio --file-num=16 --file-total-size=2G  --file-test-mode=rndrw run  --测试模式，随机读写
sysbench 0.5:  multi-threaded system evaluation benchmark

Running the test with following options:
Number of threads: 1   --线程1个
Random number generator seed is 0 and will be ignored


Threads started!

Operations performed:  6000 reads, 4000 writes, 1600 Other = 11600 Total
6000次读，4000次写（因为默认1.5:1）1600次其他操作；
Read 93.75Mb  Written 62.5Mb  Total transferred 156.25Mb  (276.57Mb/sec)
随机读了93.75M数据，写了62.5M数据 总IO 156.25M
17700.40 Requests/sec executed
每秒执行17700次
Test execution summary:
    total time:                          0.5650s 总耗费时间
    total number of events:              10000    读写次数10000次
    total time taken by event execution: 0.3961s   读写花费时间
    per-request statistics:
         min:                                  0.00ms 最小时间
         avg:                                  0.04ms 平均时间
         max:                                 20.24ms  最大时间
         approx.  95 percentile:               0.08ms  95%读写花费时间

Threads fairness:
    events (avg/stddev):           10000.0000/0.00
    execution time (avg/stddev):   0.3961/0.00

sysbench --test=fileio --file-num=16 --file-total-size=2G  --file-test-mode=rndrw cleanup 最后执行clearnup清除测试数据

 

Step 4: 线程测试：

 sysbench --num-threads=64 --test=threads --thread-yields=2000 --thread-locks=2 run 

并发线程64，采用线程测试方法，每个请求产生2000个线程，每个线程锁2个；

[root@localhost sysbench-0.5]# sysbench --test=threads --num-threads=64 --thread-yields=2000 --thread-locks=2 run
sysbench 0.5:  multi-threaded system evaluation benchmark

Running the test with following options:
Number of threads: 64  --并发线程64个
Random number generator seed is 0 and will be ignored


Threads started!


Test execution summary:
    total time:                          15.7261s   
    total number of events:              10000   
    total time taken by event execution: 1004.5724s
    per-request statistics:
         min:                                  0.58ms
         avg:                                100.46ms
         max:                               1049.29ms
         approx.  95 percentile:             246.15ms

Threads fairness:
    events (avg/stddev):           156.2500/7.99
    execution time (avg/stddev):   15.6964/0.02

Step5: 内存测试

sysbench   --num-threads=64 --test=memory --memory-block-size=8k --memory-total-size=16G --memory-scope=global --memory-access-mode=rnd run

[root@localhost sysbench-0.5]# sysbench   --num-threads=64 --test=memory --memory-block-size=8k --memory-total-size=16G --memory-scope=global --memory-access-mode=rnd run
sysbench 0.5:  multi-threaded system evaluation benchmark

Running the test with following options:
Number of threads: 64
Random number generator seed is 0 and will be ignored


Threads started!

Operations performed: 2097152 (3690132.19 ops/sec)

16384.00 MB transferred (28829.16 MB/sec)


Test execution summary:
    total time:                          0.5683s
    total number of events:              2097152
    total time taken by event execution: 20.6291s
    per-request statistics:
         min:                                  0.00ms
         avg:                                  0.01ms
         max:                                351.43ms
         approx.  95 percentile:               0.00ms

Threads fairness:
    events (avg/stddev):           32768.0000/8682.67
    execution time (avg/stddev):   0.3223/0.06

[root@localhost sysbench-0.5]#