REDO日志的基本结构

liyongdong · 发表于 2005-12-30 11:31

图片

liyongdong · 发表于 2005-12-30 11:33

假设这个参数设定为80 (Oracle7.2之前是800)，也就是说，对于80 byte以下的REDO entry由REDO allocation latch执行copy，超过80byte的REDO entry就由REDO copy latch执行copy。

liyongdong · 发表于 2005-12-30 11:35

这种结构是因为REDO allocation latch只有一个，所以当REDO allocation latch在copy比较大的REDO entry时，又发生下一个REDO
entry，就必须等到REDO allocation latch恢复闲置才能进行copy，也就是存在时间等待。因此，把copy任务交给多个REDO copy latch处理，可以预防REDO信息的copy中断。

liyongdong · 发表于 2005-12-30 11:35

此外，因为决定REDO copy latch数目是由初始设定参数LOG_SIMULTANEOUS_COPIES决定，这个参数的预设值是实际上CPU的个数，可以增加到CPU个数的2倍。(可以通过初始化参数CPU_COUNT动态指定为CPU的个数)

liyongdong · 发表于 2005-12-30 11:36

下次继续。

Yong Huang · 发表于 2005-12-30 12:02

最初由 liyongdong 发布
[B]图中提到REDO entry的大小，主要取决于初始设定参数LOG_SMALL_ENTRY_MAX_SIZE。 [/B]

Can you do some research and find out why LOG_SMALL_ENTRY_MAX_SIZE is deprecated from 8i on? I suspect the mechanism of using the copy latch when a certain size is exceeded is no longer needed. So what replaced it?

Yong Huang

netbanker · 发表于 2005-12-30 13:15

during hot backup, before image does write into redo log based on block level.

Appreciated if liyongdong and Huang Yong will discuss more of redo mechanism under some perticular situation.

liyongdong · 发表于 2005-12-31 09:48

最初由 netbanker 发布
[B]during hot backup, before image does write into redo log based on block level.

Appreciated if liyongdong and Huang Yong will discuss more of redo mechanism under some perticular situation. [/B]

perticular 应该是particular.

liyongdong · 发表于 2005-12-31 10:05

LOG_SMALL_ENTRY_MAX_SIZE and LOG_SIMULTANEOUS_COPIES are obsolete in Oracle8i.

sunnily · 发表于 2006-1-1 00:13

学习来的

REDO日志的基本结构

浏览过的版块