One of the most important metrics to address when looking at the performance of Oracle is the database buffer cache hit rate. This metric tells us how frequently Oracle found that a block that it needs is already present in the buffer cache, thus avoiding a physical I/O. The init.ora parameter DB_BLOCK_BUFFERS defines the size of the buffer cache.

Oracle does provide some help in assessing the likely impact of changes to the DB_BLOCK_BUFFERS parameter. To use this you must give another initialization parameter, DB_BLOCK_LRU_EXTENDED_STATISTICS, an integer value specifying the number of buffers that you could potentially add or remove. After a consistent period of normal database usage, the X$KCBRBH table will indicate the change in cache hits which will be caused by the addition of each buffer, up to the number specified in DB_BLOCK_LRU_EXTENDED_STATISTICS. Conversely, the X$KCBCBH table provides the same information, but for a reduction in the number of buffers.

Note - remove the DB_BLOCK_LRU_EXTENDED_STATISTICS parameter as soon as you have finished your measurement trial to avoid the overhead that it causes!!