Here's an excellent real life example of why BCHR is not a good tuning metric and you should focus on reducing I/Os.
A simple fix for a query and here is the resulting email to the client, who understands that BCHR is not good. A little techie humor...
 

I have good news and I have bad news.

The good news is that the elapsed query time and total I/Os for the latest iteration dropped significantly.

Old --> 1:42 min 2,715,659 i/os (15,925 physical)

New --> 22 seconds 3318 i/os (2861)

However, the bad news is that the Buffer Cache Hit Ratio dropped dramatically!

Old --> 99.36%

New --> 13.77%

So, I have undone all the changes I made and will begin looking at other methods to improve performance!

Reply via email to