公众号：杨建荣的学习笔记 · 作者：r3第5天 · 发布：2014-06-16 14:59:19 · 原文链接

接着续上次提到的sqlldr的性能问题，加载一个表数据400多万条记录，竟然用了6个多小时。最后大家争论不休的时候，我发现应该是网络的问题。
http://blog.itpub.net/23718752/viewspace-1182534/
今天客户IT的同事把网络做了调整，他们就想看看到底改进有多大。
下面是测试的一些记录。

àoriginal logs for issue table, loading around 6 hours.

Total logical records skipped: 0

Total logical records read: 4096786

Total logical records rejected: 0

Total logical records discarded: 0

Run began on Wed Jun 11 08:52:55 2014

Run ended on Wed Jun 11 14:57:40 2014

Elapsed time was: 06:04:44.05

CPU time was: 00:00:38.18

àafter network enhanced, loading time has reduced to around 4 mins.

Total logical records skipped: 0

Total logical records read: 4096786

Total logical records rejected: 0

Total logical records discarded: 0

Run began on Mon Jun 16 12:58:01 2014

Run ended on Mon Jun 16 13:02:15 2014

Elapsed time was: 00:04:13.28

CPU time was: 00:00:30.00

àconsidering impact from index, I loaded data again, keep table structure,index exactly the same.

Total logical records skipped: 0

Total logical records read: 4096786

Total logical records rejected: 0

Total logical records discarded: 0

Run began on Mon Jun 16 13:09:34 2014

Run ended on Mon Jun 16 13:16:37 2014

Elapsed time was: 00:07:02.61

CPU time was: 00:00:31.23

àtune sqlldr command with below options, keep table structure,index exactly the same.

parallel=true bindsize=7500000 readsize=7500000 streamsize=7500000 rows=50000

Total logical records skipped: 0

Total logical records read: 4096786

Total logical records rejected: 0

Total logical records discarded: 0

Run began on Mon Jun 16 13:21:13 2014

Run ended on Mon Jun 16 13:24:19 2014

Elapsed time was: 00:03:05.75

CPU time was: 00:00:41.27

所以综上所述，在排查了cpu,io,缓存，网络的因素后，还可以使用一些选项来优化sqlldr,在某种程度上得到提升。