用Oracle中的Statspack诊断数据库性能实例

图表：

图片看不清楚？请点击这里查看原图（大图）。

分析：

从趋势图中我们可以看出，数据库每日读操作较为平稳，数据量大约在4000左右。在下午2点到5点期

间比较繁忙。峰值达到150000左右。

数据库写操作变化也比较平稳，数据改变量在80000左右，凌晨一点半到早晨8点半左右数据库访问极少。

这是一个以写为主的数据库，我们需要更多注意的是写竞争。

2．Buffer命中率

Sql代码

select
substr(to_char(snap_time,'yyyy-mm-ddHH24:MI'),12),
round(100*(((a.value-e.value)+(b.value-f.value))-(c.value-g.value))/
((a.value-e.value)+(b.value-f.value)))
"BUFFERHITRATIO"
from
perfstat.stats$sysstata,
perfstat.stats$sysstatb,
perfstat.stats$sysstatc,
perfstat.stats$sysstatd,
perfstat.stats$sysstate,
perfstat.stats$sysstatf,
perfstat.stats$sysstatg,
perfstat.stats$snapshotsn
where
a.snap_id=sn.snap_id
and
b.snap_id=sn.snap_id
and
c.snap_id=sn.snap_id
and
d.snap_id=sn.snap_id
and
e.snap_id=sn.snap_id-1
and
f.snap_id=sn.snap_id-1
and
g.snap_id=sn.snap_id-1
and
a.statistic#=39
and
e.statistic#=39
and
b.statistic#=38
and
f.statistic#=38
and
c.statistic#=40
and
g.statistic#=40
and
d.statistic#=41
　

select
substr(to_char(snap_time,'yyyy-mm-ddHH24:MI'),12),
round(100*(((a.value-e.value)+(b.value-f.value))-(c.value-g.value))/
((a.value-e.value)+(b.value-f.value)))
"BUFFERHITRATIO"
from
perfstat.stats$sysstata,
perfstat.stats$sysstatb,
perfstat.stats$sysstatc,
perfstat.stats$sysstatd,
perfstat.stats$sysstate,
perfstat.stats$sysstatf,
perfstat.stats$sysstatg,
perfstat.stats$snapshotsn
where
a.snap_id=sn.snap_id
and
b.snap_id=sn.snap_id
and
c.snap_id=sn.snap_id
and
d.snap_id=sn.snap_id
and
e.snap_id=sn.snap_id-1
and
f.snap_id=sn.snap_id-1
and
g.snap_id=sn.snap_id-1
and
a.statistic#=39
and
e.statistic#=39
and
b.statistic#=38
and
f.statistic#=38
and
c.statistic#=40
and
g.statistic#=40
and
d.statistic#=41