1、sysbench测试主机在内存中的连续读写速度
2、测试环境，虚拟机4线程、1G内存，PC服务器96线程、512G内存
虚拟机传输16G数据、块大小16K、线程数4

root@pm1:/root# sysbench --test=memory --memory-block-size=16K --memory-total-size=16G --num-threads=4 run
sysbench 0.5: multi-threaded system evaluation benchmark
Running the test with following options:
Number of threads: 4
Random number generator seed is 0 and will be ignored
Threads
Operations performed: 1048576 (269059.42 ops/sec)
16384.00 MB transferred (4204.05 MB/sec)
General statistics:
total time: 3.8972s
total number of events: 1048576
total time taken by event execution: 9.0149s
response time:
min: 0.00ms
avg: 0.01ms
max: 29.98ms
approx. 95 percentile: 0.01ms
Threads fairness:
events (avg/stddev): 262144.0000/2079.32
execution time (avg/stddev): 2.2537/0.02

PC服务器传输16G、块大小16K、线程数96

[root@ ~]# sysbench --test=memory --memory-block-size=16K --memory-total-size=16G --num-threads=96 run
sysbench 0.5: multi-threaded system evaluation benchmark
Running the test with following options:
Number of threads: 96
Random number generator seed is 0 and will be ignored
Threads
Operations performed: 1048576 (405615.58 ops/sec)
16384.00 MB transferred (6337.74 MB/sec)
General statistics:
total time: 2.5851s
total number of events: 1048576
total time taken by event execution: 47.7844s
response time:
min: 0.00ms
avg: 0.05ms
max: 8.60ms
approx. 95 percentile: 0.08ms
Threads fairness:
events (avg/stddev): 10922.6667/408.97
execution time (avg/stddev): 0.4978/0.02

看来多线程在内存中的调度还是很大优势的，如果把PC服务器换成4线程，其实和虚拟机性能差不多。

[root@ ~]# sysbench --test=memory --memory-block-size=16K --memory-total-size=16G --num-threads=4 run
sysbench 0.5: multi-threaded system evaluation benchmark
Running the test with following options:
Number of threads: 4
Random number generator seed is 0 and will be ignored
Threads
Operations performed: 1048576 (271394.71 ops/sec)
16384.00 MB transferred (4240.54 MB/sec)
General statistics:
total time: 3.8637s
total number of events: 1048576
total time taken by event execution: 13.2885s
response time:
min: 0.00ms
avg: 0.01ms
max: 0.19ms
approx. 95 percentile: 0.04ms
Threads fairness:
events (avg/stddev): 262144.0000/11704.01
execution time (avg/stddev): 3.3221/0.03
You have new mail in /var/spool/mail/root

上面是把PC服务器的线程数改成4之后，每秒传输4240M，跟之前虚拟机的传输差不多，但是在最大响应时间差别比较大。

ITPub博客

测试软件：sysbench-memory