Regardless of Oracle's LRU algorithm, every Oracle LRU chain has a least recently used (LRU) end. It also has a most recently used (MRU) end. In very general terms, the more popular buffer headers will reside near the MRU end, and the less popular buffers will reside near the LRU end.

The standard LRU algorithm is very simple. When a buffer is brought into the cache or accessed (query or DML), it is placed at the MRU end of the session's LRU chain (each session is associated with an LRU chain). The thinking is that a popular buffer will be repeatedly touched and therefore repeatedly moved to the MRU end of the LRU chain. The movement to the MRU of this list is commonly called, buffer promotion. If a buffer is not popular, then as other buffers are promoted or inserted into the LRU, the unpopular buffers will naturally migrate toward the LRU end of the LRU chain.

Possibly lurking near the LRU end of every LRU chain is a server process seeking a free unpopular buffer to replace with a block it just read from disk. I call such a server process the reaper, as it replaces buffers whose goal in life is to live long and gloriously in Oracle's buffer cache. For example, suppose the LRU situation is as shown in Figure 6-16 and there is a server process looking for a free buffer. The server process will start at the LRU end of the LRU chain and check if buffer header BH 310 is free. If buffer header BH 310 indicates it is free, it will be replaced (being transported to buffer heaven, we hope) with the just-read block's buffer header and moved to the MRU end of the list.