sysbench 基准测试简介

来源：互联网发布：淘宝超级店长软件编辑：程序博客网时间：2024/05/16 08:14

1.简介

sysbench是一款开源的多线程性能测试工具，可以执行CPU/内存/线程/IO/数据库等方面的性能测试。数据库目前支持MySQL/Oracle/PostgreSQL。本文只是简单演示一下几种测试的用法，后续准备利用sysbench来对MySQL进行一系列的测试。具体的一些参数设置，需要根据不同的测试要求来进行调整。

2.安装

#安装yum包及配置环境变量yum -y install bzr libtool wget http://imysql.com/wp-content/uploads/2014/09/sysbench-0.4.12-1.1.tgzecho 'export LD_LIBRARY_PATH=/Data/apps/mysql/lib' >> /etc/profilesource /etc/profile# 安装sysbenchmkdir -p /Data/apps/sysbenchmkdir -p /Data/apps/sysbench/logtar -zxvf sysbench-0.4.12-1.1.tgzcd sysbench-0.4.12-1.1./autogen.sh./configure --prefix=/Data/apps/sysbench --with-mysql-includes=/Data/apps/mysql/include --with-mysql-libs=/Data/apps/mysql/libmake && make installcp /Data/apps/sysbench/bin/sysbench /usr/local/bin/# /Data/package/sysbench-0.4.12-1.1为安装包目录cp -a /Data/package/sysbench-0.4.12-1.1/sysbench/tests /Data/apps/sysbench/

3.测试

3.1 Mysql数据库测试

mysql > create database sbtest;
    sysbench --mysql-host=127.0.0.1 --mysql-port=3306 --mysql-user=root --mysql-password='' --test=/Data/apps/sysbench/tests/db/oltp.lua --oltp_tables_count=10 --oltp-table-size=100000 --rand-init=on prepare

# --test=tests/db/oltp.lua 表示调用 tests/db/oltp.lua 脚本进行 oltp 模式测试
# --oltp_tables_count=10 表示会生成 10 个测试表
# --oltp-table-size=100000 表示每个测试表数据行数
# --rand-init=on 表示每个测试表都是用随机数据来填充的

sysbench --mysql-host=127.0.0.1 --mysql-port=3306 --mysql-user=root --mysql-password='' --test=/Data/apps/sysbench/tests/db/oltp.lua --oltp_tables_count=10 --oltp-table-size=10000000 --num-threads=8 --oltp-read-only=off --report-interval=10 --rand-type=uniform --max-time=120 --max-requests=0 --percentile=99 run >> /Data/apps/sysbench/log/sysbench_oltp.log

# --num-threads=8 表示发起 8个并发连接
# --oltp-read-only=off 表示不要进行只读测试，也就是会采用读写混合模式测试
# --report-interval=10 表示每10秒输出一次测试进度报告
# --rand-type=uniform 表示随机类型为固定模式，其他几个可选随机模式：uniform(固定),gaussian(高斯),special(特定的),pareto(帕累托)
# --max-time=120 表示最大执行时长为 120秒
# --max-requests=0 表示总请求数为 0，因为上面已经定义了总执行时长，所以总请求数可以设定为 0；也可以只设定总请求数，不设定最大执行时长
# --percentile=99 表示设定采样比例，默认是 95%，即丢弃1%的长请求，在剩余的99%里取最大值

==========================测试结果解读如下==========================sysbench 0.5:  multi-threaded system evaluation benchmarkRunning the test with following options:Number of threads: 8Report intermediate results every 10 second(s)Random number generator seed is 0 and will be ignoredThreads started![  10s] threads: 8, tps: 26.60, reads/s: 410.22, writes/s: 115.11, response time: 1050.29ms (99%)[  20s] threads: 8, tps: 32.40, reads/s: 462.00, writes/s: 131.60, response time: 699.68ms (99%)[  30s] threads: 8, tps: 33.00, reads/s: 467.60, writes/s: 133.20, response time: 496.50ms (99%)[  40s] threads: 8, tps: 30.80, reads/s: 441.00, writes/s: 125.30, response time: 617.02ms (99%)[  50s] threads: 8, tps: 35.80, reads/s: 504.00, writes/s: 143.80, response time: 526.03ms (99%)[  60s] threads: 8, tps: 33.20, reads/s: 468.50, writes/s: 133.30, response time: 774.87ms (99%)                                        .........[3540s] threads: 8, tps: 41.90, reads/s: 602.00, writes/s: 170.90, response time: 483.88ms (99%)[3550s] threads: 8, tps: 43.40, reads/s: 616.00, writes/s: 175.40, response time: 383.69ms (99%)[3560s] threads: 8, tps: 38.00, reads/s: 543.20, writes/s: 154.40, response time: 475.41ms (99%)[3570s] threads: 8, tps: 39.60, reads/s: 560.00, writes/s: 159.60, response time: 425.82ms (99%)[3580s] threads: 8, tps: 42.10, reads/s: 600.60, writes/s: 170.80, response time: 384.04ms (99%)[3590s] threads: 8, tps: 38.10, reads/s: 547.40, writes/s: 155.40, response time: 542.17ms (99%)[3600s] threads: 8, tps: 40.40, reads/s: 574.00, writes/s: 163.40, response time: 459.05ms (99%)OLTP test statistics:    queries performed:        read:                            2068584    -- 读总数        write:                           588759      -- 写总数        other:                           293247-- 其他操作总数(SELECT、INSERT、UPDATE、DELETE之外的操作，例如COMMIT等)        total:                           2950590-- 全部总数    transactions:                        145491 (40.41 per sec.)-- 总事务数(每秒事务数)    deadlocks:                           2265   (0.63 per sec.)-- 发生死锁总数    read/write requests:                 2657343 (738.12 per sec.)-- 读写总数(每秒读写次数)    other operations:                    293247 (81.45 per sec.)-- 其他操作总数(每秒其他操作次数)General statistics:    total time:                          3600.1452s-- 总耗时    total number of events:              145491-- 共发生多少事务数    total time taken by event execution: 28800.5977s-- 所有事务耗时相加(不考虑并行因素)    response time:         min:                                 49.84ms-- 最小耗时         avg:                                197.95ms-- 平均耗时         max:                               1355.40ms-- 最长耗时         approx.  99 percentile:             433.54ms-- 超过99%平均耗时Threads fairness:    events (avg/stddev):           18186.3750/63.14    execution time (avg/stddev):   3600.0747/0.04

3.2 CPU测试

sysbench CPU测试使用64位整数，测试计算素数直到某个最大值所需要的时间

shell> sysbench --test=cpu --cpu-max-prime=2000 runsysbench 0.5:  multi-threaded system evaluation benchmarkRunning the test with following options:Number of threads: 1Random number generator seed is 0 and will be ignoredPrimer numbers limit: 2000Threads started!Test execution summary:    total time:                          2.2452s    total number of events:              10000    total time taken by event execution: 2.2347s    per-request statistics:         min:                                  0.20ms         avg:                                  0.22ms         max:                                  3.35ms         approx.  95 percentile:               0.27msThreads fairness:    events (avg/stddev):           10000.0000/0.00    execution time (avg/stddev):   2.2347/0.00补充：查看CPU信息方法查看物理cpu个数 grep "physical id" /proc/cpuinfo | sort -u | wc -l查看核心数量 grep "core id" /proc/cpuinfo | sort -u | wc -l查看线程数量 grep "processor" /proc/cpuinfo | sort -u | wc -l在sysbench的测试中，--num-threads取值为"线程数量"即可

3.3 线程(thread)测试

测试线程调度器的性能。对于高负载情况下测试线程调度器的行为非常有用

shell> sysbench --test=threads --num-threads=64 --thread-yields=100 --thread-locks=2 runsysbench 0.5:  multi-threaded system evaluation benchmarkRunning the test with following options:Number of threads: 64Random number generator seed is 0 and will be ignoredThreads started!Test execution summary:    total time:                          1.9581s    total number of events:              10000    total time taken by event execution: 124.8938s    per-request statistics:         min:                                  0.05ms         avg:                                 12.49ms         max:                                151.15ms         approx.  95 percentile:              50.83msThreads fairness:    events (avg/stddev):           156.2500/14.48    execution time (avg/stddev):   1.9515/0.00

3.4 文件IO性能测试

生成需要的测试文件，文件总大小5G，16个并发线程。执行完后会在当前目录下生成一堆小文件。

shell> sysbench --test=fileio --num-threads=16 --file-total-size=5G prepare

执行测试，指定随机读写模式

指定读写模式：

seqwr 顺序写入
seqrewr 顺序重写
seqrd 顺序读取
rndrd 随机读取
rndwr 随机写入
rndrw 混合随机读/写

shell> sysbench --test=fileio --num-threads=16 --init-rng=on --file-total-size=5G --file-test-mode=rndrw runsysbench 0.5:  multi-threaded system evaluation benchmarkRunning the test with following options:Number of threads: 16Random number generator seed is 0 and will be ignoredThreads started!Operations performed:  5999 reads, 4001 writes, 12800 Other = 22800 TotalRead 93.734Mb  Written 62.516Mb  Total transferred 156.25Mb  (9.2561Mb/sec)   ##吞吐量  592.39 Requests/sec executedTest execution summary:    total time:                          16.8808s    total number of events:              10000    total time taken by event execution: 176.1816s    per-request statistics:         min:                                  0.01ms         avg:                                 17.62ms         max:                                416.73ms         approx.  95 percentile:             104.82msThreads fairness:    events (avg/stddev):           625.0000/62.39    execution time (avg/stddev):   11.0114/0.67

清除测试文件

shell> sysbench --test=fileio --num-threads=16 --file-total-size=5G cleanup

3.5 互斥锁(Mutex)测试

测试互斥锁的性能，方式是模拟所有线程在同一时刻并发运行，并都短暂请求互斥锁。

shell> sysbench --test=mutex --num-threads=16 --mutex-num=1024 --mutex-locks=10000 --mutex-loops=5000 runsysbench 0.5:  multi-threaded system evaluation benchmarkRunning the test with following options:Number of threads: 16Random number generator seed is 0 and will be ignoredThreads started!Test execution summary:    total time:                          0.0135s    total number of events:              16    total time taken by event execution: 0.0411s    per-request statistics:         min:                                  0.70ms         avg:                                  2.57ms         max:                                  9.19ms         approx.  95 percentile:               9.16msThreads fairness:    events (avg/stddev):           1.0000/0.00    execution time (avg/stddev):   0.0026/0.00

3.6 内存测试

内存测试测试了内存的连续读写性能。

shell> sysbench --test=memory --num-threads=16 --memory-block-size=8192 --memory-total-size=1G runsysbench 0.5:  multi-threaded system evaluation benchmarkRunning the test with following options:Number of threads: 16Random number generator seed is 0 and will be ignoredThreads started!Operations performed: 131072 (381158.38 ops/sec)1024.00 MB transferred (2977.80 MB/sec)       ##吞吐量Test execution summary:    total time:                          0.3439s    total number of events:              131072    total time taken by event execution: 3.9915s    per-request statistics:         min:                                  0.00ms         avg:                                  0.03ms         max:                                 51.02ms         approx.  95 percentile:               0.00ms  ##大约95%的时间分布Threads fairness:    events (avg/stddev):           8192.0000/1166.77    execution time (avg/stddev):   0.2495/0.02

《sysbench安装、使用、结果解读》

《sysbench 0.5 基准测试》

0 0