MySQL内存表、MyISAM表、MemCache实现Session效率比较

早先问过BleakWind,他认为,给别人做的话MySQL内存表做会话比较好一些,因为MySQL内存表做Session更容易维护(可以制作安装脚本)。这个周末,我进行了一些测试,测试MySQL MyISAM表做会话(对时间给不给索引)、内存表做会话、MemCache做会话的效率比较。
定义会话Session类:

定义会话处理器接口(Session_Handler_Interface):

实现MySQL内存表Session处理器:

实现MemCache会话处理器:

MySQL内存表测试程序:

MemCache会话测试代码:

MySQL会话表如下:
CREATE TABLE `session` (
`id` char(32) COLLATE utf8_unicode_ci NOT NULL,
`name` varchar(20) COLLATE utf8_unicode_ci NOT NULL,
`ip` varchar(15) COLLATE utf8_unicode_ci NOT NULL,
`time` int(10) unsigned NOT NULL,
PRIMARY KEY (`id`)
) ENGINE=MEMORY DEFAULT CHARSET=utf8 COLLATE=utf8_unicode_ci;
测试结果:
MemCache:
======================================================================================
Concurrency Level:      30
Time taken for tests:   18.234375 seconds
Complete requests:      10000
Failed requests:        0
Write errors:           0
Total transferred:      2774781 bytes
HTML transferred:       280000 bytes
Requests per second:    548.41 [#/sec] (mean)
Time per request:       54.703 [ms] (mean)
Time per request:       1.823 [ms] (mean, across all concurrent requests)
Transfer rate:          148.57 [Kbytes/sec] received
Connection Times (ms)
              min mean[+/-sd] median   max
Connect:        0    0   1.9      0      31
Processing:     0   53 12.1     46     328
Waiting:        0   52 11.9     46     328
Total:          0   53 12.1     46     328
Percentage of the requests served within a certain time (ms)
50%     46
66%     62
75%     62
80%     62
90%     62
95%     62
98%     62
99%     78
100%    328 (longest request)
MySQL内存表:
=================================================================================
Concurrency Level:      30
Time taken for tests:   20.375000 seconds
Complete requests:      10000
Failed requests:        4694
   (Connect: 0, Length: 4694, Exceptions: 0)
Write errors:           0
Total transferred:      3118440 bytes
HTML transferred:       623048 bytes
Requests per second:    490.80 [#/sec] (mean)
Time per request:       61.125 [ms] (mean)
Time per request:       2.038 [ms] (mean, across all concurrent requests)
Transfer rate:          149.45 [Kbytes/sec] received
Connection Times (ms)
              min mean[+/-sd] median   max
Connect:        0    0   1.9      0      31
Processing:     0   60   7.6     62     125
Waiting:        0   59   7.6     62     125
Total:          0   60   7.5     62     125
Percentage of the requests served within a certain time (ms)
50%     62
66%     62
75%     62
80%     62
90%     62
95%     62
98%     78
99%     78
100%    125 (longest request)
其他测试:
将MySQL实现的数据库引擎改为 MyISAM(依然为 并发 30 测试 10000 次) 结果为:
Requests per second:    440.17 [#/sec] (mean)
Percentage of the requests served within a certain time (ms)
50%     62
66%     62
75%     78
80%     78
90%     78
95%     78
98%     78
99%     78
100%    640 (longest request)
为MyISAM表的 time 列增加索引(因为 会话表 读写次数几乎相等,因此应该效果不明显),结果为:
Requests per second:    441.08 [#/sec] (mean)
Percentage of the requests served within a certain time (ms)
50%     62
66%     62
75%     78
80%     78
90%     78
95%     78
98%    109
99%    109
100%    156 (longest request)
=================================================================================
结论:
     MemCache做会话效率最高,也最灵活,但目前尝不支持查看谁在线等功能,附加的,只能自己增加一个数组记录在线用户、以及最后活跃时间并实现gc等。
     MySQL内存表做会话效率也相当的高,另外一个有点是,MySQL内存表似乎更稳定,longest request (125ms)明显的短于 MemCache的(328ms)。不过缺点是,存储的字段数以及字段长度受限。
     MySQL MyISAM表做会话在这里居然也达到了440的rps,真不简单。不过您要是等半个小时在测试一次,您就会明白MyISAM表的缺点了,页面几乎崩溃。MyISAM表的缺点是,一旦其中有较多的碎片,这个数据库几乎都不可用了,您注释掉 gc 的代码在这里貌似可以获得更好的效率表现(^_^、当然也可以定时的Optimizer,概率触发或者Cron定时启动也不错)
     MyISAM表对time列增加索引对每秒完成的请求数没什么影响,不过有一点需要注意的是,增加索引后,每次完成 request的时间更均匀了,longest request从640ms跌到了156ms。为time列增加索引有助于让完成请求的时间变均匀。
测试平台:
Acer Aspire 4520G:
CPU:AMD Athlon 64 * 2 TK-55
RAM: IG
OS: Windows XP sp2
Web Server: Apache 2.26
PHP: PHP5.25