1.sysbench安装：
yum install sysbench
2.测试cpu的性能指标
第一次：sysbench cpu run表示使用一个thread执行一次cpu性能测试;测试结果如下:
sysbench 1.0.17 (using system LuaJIT 2.0.4) //sysbench的版本
Running the test with following options:
Number of threads: 1 //使用单个线程处理
Initializing random number generator from current time
Prime numbers limit: 10000 //最多产生1000个素数
Initializing worker threads...
Threads started!
CPU speed:
events per second: 711.74 //每秒可以处理711事件
General statistics:
total time: 10.0013s //总共花费的时间（秒）
total number of events: 7120 //共生产的事件数
Latency (ms):
min: 1.36 //每个事件最小处理时间(毫秒)
avg: 1.40 //平均每个事件的处理时间
max: 3.38 //最大处理时间
95th percentile: 1.55 //95%的事件的处理时间
sum: 9988.65 //总共的处理时间
Threads fairness:
events (avg/stddev):7120.0000/0.00 //每个线程处理的总事件；标准方差（单线程无意义）
execution time (avg/stddev): 9.9887/0.00 //每个线程处理的总时间，标准方差(单线程无意义)
第二次：sysbench --threads=2 cpu run 使用两个线程测试cpu
sysbench 1.0.17 (using system LuaJIT 2.0.4)
Running the test with following options:
Number of threads: 2 //两个线程
Initializing random number generator from current time
Prime numbers limit: 10000
Initializing worker threads...
Threads started!
CPU speed:
events per second: 1399.61 //每秒处理1399.61个事件
General statistics:
total time: 10.0011s //一共执行了10秒
total number of events: 14001 //共产生14001个事件
Latency (ms):
min: 1.28 //事件执行的最小时间（毫秒）
avg: 1.43 //平均执行时间
max: 4.16 //最大执行时间
95th percentile: 1.64 //95%事件均在1.64毫秒内完成
sum:19975.20
Threads fairness:
events (avg/stddev):7000.5000/12.50 //两个线程的标准方差为12.50，该值越小，说明越稳定
execution time (avg/stddev): 9.9876/0.00 //每个线程平均执行总时间
通过第1次与第2次的结果来看，增加一个线程基本可以多一倍的处理能力；我们继续增加线程
第三次，执行sysbench --threads=3 cpu run
sysbench 1.0.17 (using system LuaJIT 2.0.4)
Running the test with following options:
Number of threads: 3
Initializing random number generator from current time
Prime numbers limit: 10000
Initializing worker threads...
Threads started!
CPU speed:
events per second: 1217.41
General statistics:
total time: 10.0007s
total number of events: 12178
Latency (ms):
min: 1.56
avg: 2.46
max: 14.25
95th percentile: 7.56
sum:29947.76
Threads fairness:
events (avg/stddev):4059.3333/385.33
execution time (avg/stddev): 9.9826/0.00
通过第2次与第3次的结果相对比，发现多一个线程后每秒完成的事件稍有下降，并没有提升；我们继续增加一个线程进行第4次测试；
sysbench 1.0.17 (using system LuaJIT 2.0.4)
Running the test with following options:
Number of threads: 4
Initializing random number generator from current time
Prime numbers limit: 10000
Initializing worker threads...
Threads started!
CPU speed:
events per second: 1391.58
General statistics:
total time: 10.0017s
total number of events: 13921
Latency (ms):
min: 1.34
avg: 2.87
max: 15.37
95th percentile: 7.84
sum:39916.69
Threads fairness:
events (avg/stddev):3480.2500/52.34
execution time (avg/stddev): 9.9792/0.01
4个线程的处理结果与2个3个的线程处理结果基本处于一样的处理结果，并没有明显的提升，因此这时可以基本确定,本台测试服务器的cpu线程数应该是2个；为了验证是否准确，我们查看一下当前服务器的cpu情况：
cat /proc/cpuinfo | grep "physical id" | sort | uniq | wc -l
通过结果查看，验证了我们通过sysbench的测试结果。
如果觉得《性能测试sysbench之cpu性能测试》对你有帮助，请点赞、收藏，并留下你的观点哦！
