Figure 8-3. Shown is an abstracted redo log buffer, indicating the two pointers: the last redo allocation pointer and the last redo write pointer. The various markers (MK...) are used in describing specific scenarios in the text.

Suppose a server process needs redo related to a buffer change. It creates the change vectors and combines them into a single redo entry. At this point, the server process knows how many bytes of redo are required.

Oracle9i Release 2 introduced significant changes in the redo log buffer architecture and algorithms. Yet, even prior to this release, there are adjustments the performance specialist can make to alter how Oracle's redo mechanism operates. Let's look at how the redo log buffer operates and some possible adjustments in both pre- and post-Oracle9i Release 2 versions.