Buffer Cache Hit Ratio Is Not Everything

[@more@]

For example, suppose two applications (A and B) return the same result set. Application A
has a cache hit ratio of 99%, whereas Application B has a hit ratio of 60 %. Which
application has the best performance? On further investigation you notice that Application
A requires 100,000 logical reads and 1,000 physical reads, whereas Application B has 100
logical reads and 40 physical reads. Which application is better tuned now? With the
additional information, the answer is B. There is clearly something wrong with
Application A, when it does so many logical reads to return a small result set.

从上面的example来看，官方给出的buffer cache命中率只不过是调优数据库性能的一部分，而不是全部。

多年的DBA工作经验告诉我，的确如此。