从MySQL的源码剖析Innodb buffer的命中率计算

按官方手册推荐innodb buffer hit ratios的计算是:
 

100-((ireads / ireadrequests)*100)
ireads : mysql->status->innodb_buffer_pool_reads
ireadrequests: mysql->status->innodb_buffer_pool_read_requests

出处: http://dev.mysql.com/doc/mysql-monitor/2.0/en/mem_graphref.html
搜”hit ratios”
推荐有兴趣的同学把这个页面都看一下应该也会有很大收获.
另外在hackmysql: www.hackmysql.com网站上的: mysqlsqlreport中关于buffer命中计算是:

 

$ib_bp_read_ratio = sprintf "%.2f",
($stats{'innodb_buffer_pool_read_requests'} ?
100 - ($stats{'innodb_buffer_pool_reads'} /
$stats{'innodb_buffer_pool_read_requests'}) * 100 :0);

即: 

ib_bp_hit=100-(innodb_buffer_pool_reads/innodb_buffer_pool_read_requests)*100

另外我们知道查看innodb buffer hit ratios的地方是:

show engine innodb status\g;

buffer pool hit rate : xxxx/1000;
那个xxx/1000即是buffer pool hit ratios的命中.
这样也可以从代码里看一下这个bp命中计算:

storage/innobase/buf/buf0buf.c # void buf_print_io
storage/innodbase/include/buf0buf.h #struct buf_block_struct

在buf0buf.c 中的buf_print_io函数中可以看到:
 

void
buf_print_io(
…
 
if (buf_pool->n_page_gets > buf_pool->n_page_gets_old) {
fprintf(file, "buffer pool hit rate %lu / 1000\n",
(ulong)
(1000 - ((1000 * (buf_pool->n_pages_read
- buf_pool->n_pages_read_old))
/ (buf_pool->n_page_gets
- buf_pool->n_page_gets_old))));
} else {
fputs("no buffer pool page gets since the last printout\n",
file);
}
 
buf_pool->n_page_gets_old = buf_pool->n_page_gets;
buf_pool->n_pages_read_old = buf_pool->n_pages_read;
…
}

结合:
storage\innobase\include\buf0buf.h中

struct buf_block_struct{
…
ulint n_pages_read; /* number read operations */
…
ulint n_page_gets; /* number of page gets performed;
also successful searches through
the adaptive hash index are
counted as page gets; this field
is not protected by the buffer
pool mutex */
…
ulint n_page_gets_old;/* n_page_gets when buf_print was
last time called: used to calculate
hit rate */
…
ulint n_pages_read_old;/* n_pages_read when buf_print was
last time called */
…

 

从这个来看innodb buffer hit ratios的命中计算需要本次取的值和上次值做一个减法公式应该为

ib_bp_hit=1000 – (t2.ireads – t1.ireads)/(t2.ireadrequest – t1.ireadrequest)*1000

t(n): 时间点 两个时间间隔最少是30秒以上,在小意义不大.

ireads: innodb_buffer_pool_reads
ireadrequest: innodb_buffer_pool_read_requests

对innodb的输出参数有兴趣的可以关注: storage/innobase/buf/srv0srv.c 中的:

void srv_export_innodb_status()

思考:
对于innodb_buffer_pool_read_requests, innodb_buffer_pool_reads这种累加值,当很大时进行: innodb_buffer_pool_reads/innodb_buffer_pool_read_requests 相来讲只能得到从开始到现在的命中率的表现了. 如果想得到现在近五分钟,近一分钟或是8点到9点每分钟的命中率情况,如果还是按着innodb_buffer_pool_reads/innodb_buffer_pool_read_requests 进行计算,只能得到mysqld开起累计在8点-9点的每分钟的累计平均命中情况.
所以如果想到每(五)分钟的命中情况,就需要本次取得的值和一(五)分钟前的值进行相减,然后进行运算.这样才能得到一个当下的bp命中情况.
两种方法没实质的对错的问题,但相对于源码中的那种计算方式更容让发现数据库的抖动问题.

能解决的问题:
偶而的数据库性能抖动能直观的反应出来.