Sizing the Redo Log Buffer

Imagine you’ve inherited an existing Oracle database that is used to support
your company’s order-taking system. After examining the Shared Pool
and Database Buffer Cache hit ratios, you realize that they could probably
be improved by making these structures larger, but there is not sufficient
memory in the server to do so. The Redo Log Buffer’s Retry Ratio is low—
0.0000012.
Before you give up and ask the boss to order more memory for the server,
examine the size of the Redo Log Buffer. Chances are, with a Retry Ratio this
low, the Redo Log Buffer may be oversized and wasting memory.
Suppose you check the size of the Redo Log Buffer and discover that it is
10MB. Any value above 1MB will be largely unused because LGWR is signaled
to write when the Redo Log Buffer is one-third full or filled with 1MB
of redo entries. By reducing the size of the Redo Log Buffer to 1MB or less,
you can allocate the approximately 9MB remaining space to the Shared
Pool and/or Database Buffer Cache. This will likely improve the overall SGA
hit ratios without consuming any additional memory on the server.