- UID
- 22945
- 阅读权限
- 20
- 帖子
- 715
- 精华贴数
- 1
- 技术排名
- 2490
- 技术积分
- 1082
- 社区排名
- 3143
- 社区积分
- 265
- 注册时间
- 2002-9-17
- 精华贴数
- 1
- 技术积分
- 1082
- 社区积分
- 265
- 注册时间
- 2002-9-17
- 论坛徽章:
- 5
|
发表于 2005-12-15 14:45:10
|显示全部楼层
最初由 Yong Huang 发布
[B]
The gist of "Max(db_file_multiblock_read_count) *db_block_size= MaxOsIOsize" is really that you can't arbitrarily increase db_file_multiblock_read_count and expect Oracle to fetch data in db_file_multiblock_read_count*db_block_size byte chunks. I.e., there's an actual upper limit, above which the I/O size is fixed, regardless the setting of db_file_multiblock_read_count. That upper limit is set by OS. Whether changing MaxOsIOsize to MaxOracleIOsize is immaterial.
On the other hand, we need to clarify the meaning of MaxOsIOsize in our discussion. Suppose on Solaris you set maxphys to 512k now and in fact you could set it higher up to 1M (beyond which you get an error on reboot). Is 512k or 1M the MaxOsIOsize you're talking about?
Yong Huang [/B]
实际上我的意思很简单,就是OS的IO不一定等于ORACLE IO,
而Max(db_file_multiblock_read_count) *db_block_size实际上是反映的ORACLE的I/O。我记得TOM也说过,对于ORACLE来说,最小的IO单位是一个BLOCK。
另外,ORACLE的BLOCK大小是OS BLOCK的整数倍,而OS的 IO大小是OS 的文件系统的BLOCK大小,所以说我觉得ORALCE的IO大小要大于OS的IO的大小,因为它相对于OS 的IO来说,是“逻辑”的了。
另外我对Solaris平台不熟悉,maxphys这个参数不知道确切的含义 |
|