Memcached和Mysql数据存储测试对比
昨晚写了个关于Memcached和Mysql的数据存储与读取的数据对比，然后让服务器跑了一晚上，一直担心把电脑跑崩溃，还好没有，测试结果和预测的差不多。
测试的平台是Memcached-1.4.13和MySql-14.14；
访问数据库工具是iBatIS+JDBC；
操作系统CentOS6；
CPU：Pentium(R) Dual-Core CPU E5300 @ 2.60GHz双核；
网络环境：本地自环回路；
测试线程100条，每条线程分别进行100次读写，统计平均时间和总时间。
最后的结论是Memcached的单独删除操作平均耗时是5~6ms，单独写操作平均耗时是2~3.5ms，单独读操作平均耗时是1.5~3ms，可见读操作和写操作的耗时基本上是一样的，这跟Memcached内部采用Hash查找算法有关吧，而删除操作是写和读的两倍的原因是删除操作要先查找再写，所以时间消耗是其两倍。
而对Mysql的测试没有做关于delete操作和update操作的统计，只对insert和select操作进行了统计，而且数据表中的操作只有两个键，项越多，数据库越复杂，耗时应该更大。统计结果表明，向空数据表中写100条数据的平均耗时是39ms（一条一条的写，非批量写），向含有10240（10KB）条记录的数据库中读数据的情况为：单独读100条数据的平均耗时为：15ms~25ms，而批量读数据（一次读100条记录）的耗时是50~150ms，平均到每条记录的读取时间是0.5~1.5ms。
心得体会：在进行数据库读数据的时候如果数据比较集中，建议先批量读出来存放到缓冲区，现在很多数据库访问工具都集成了对缓冲区的管理。在进行数据的分散读写操作的时候建议先采用缓冲操作，再进行对数据库的相应操作，使用本地内存缓冲的速度应该比Memcached更快。
下面附录上具体的测试结果，免得以后想用的时候又找不到了，我一向的收拾不好，而且忘性还很大，呵呵。
thread 40: runs: 100 stores of obj 10KB — avg time per req 3.64 ms (total: 364 ms)
thread 40: runs: 100 gets of obj 10KB — avg time per req 1.8 ms (total: 180 ms)
thread 41: runs: 100 deletes of obj 10KB — avg time per req 5.1 ms (total: 510 ms)
thread 41: runs: 100 stores of obj 10KB — avg time per req 3.33 ms (total: 333 ms)
thread 41: runs: 100 gets of obj 10KB — avg time per req 2.29 ms (total: 229 ms)
thread 42: runs: 100 deletes of obj 10KB — avg time per req 5.28 ms (total: 528 ms)
thread 42: runs: 100 stores of obj 10KB — avg time per req 3.64 ms (total: 364 ms)
thread 42: runs: 100 gets of obj 10KB — avg time per req 1.83 ms (total: 183 ms)
thread 43: runs: 100 deletes of obj 10KB — avg time per req 5.65 ms (total: 565 ms)
thread 43: runs: 100 stores of obj 10KB — avg time per req 3.29 ms (total: 329 ms)
thread 43: runs: 100 gets of obj 10KB — avg time per req 1.65 ms (total: 165 ms)
thread 44: runs: 100 deletes of obj 10KB — avg time per req 5.94 ms (total: 594 ms)
thread 44: runs: 100 stores of obj 10KB — avg time per req 2.98 ms (total: 298 ms)
thread 44: runs: 100 gets of obj 10KB — avg time per req 1.65 ms (total: 165 ms)
thread 45: runs: 100 deletes of obj 10KB — avg time per req 5.42 ms (total: 542 ms)
thread 45: runs: 100 stores of obj 10KB — avg time per req 2.93 ms (total: 293 ms)
thread 45: runs: 100 gets of obj 10KB — avg time per req 1.8 ms (total: 180 ms)
thread 46: runs: 100 deletes of obj 10KB — avg time per req 6.05 ms (total: 605 ms)
thread 46: runs: 100 stores of obj 10KB — avg time per req 3.43 ms (total: 343 ms)
thread 46: runs: 100 gets of obj 10KB — avg time per req 1.38 ms (total: 138 ms)
thread 47: runs: 100 deletes of obj 10KB — avg time per req 5.36 ms (total: 536 ms)
thread 47: runs: 100 stores of obj 10KB — avg time per req 3.06 ms (total: 306 ms)
thread 47: runs: 100 gets of obj 10KB — avg time per req 2.22 ms (total: 222 ms)
thread 48: runs: 100 deletes of obj 10KB — avg time per req 6.29 ms (total: 629 ms)
thread 48: runs: 100 stores of obj 10KB — avg time per req 2.62 ms (total: 262 ms)
thread 48: runs: 100 gets of obj 10KB — avg time per req 1.62 ms (total: 162 ms)
thread 49: runs: 100 deletes of obj 10KB — avg time per req 4.09 ms (total: 409 ms)
thread 49: runs: 100 stores of obj 10KB — avg time per req 3.34 ms (total: 334 ms)
thread 49: runs: 100 gets of obj 10KB — avg time per req 2.62 ms (total: 262 ms)
thread 50: runs: 100 deletes of obj 10KB — avg time per req 5.56 ms
