Re: buffer hit cache ratio

"Geomancer" <pharfromhome_at_hotmail.com> wrote in message news:cf90fb89.0312221936.14869401_at_posting.google.com...
> > Tuning by the BCHR is not done by today's top Oracle professionals.
>
> Depends on who you consider a "top" professional.
>
> Tom Kyte and Cary Millsap both argue that the BHR is not a panacea,
> but is IS useful for detecting when the data buffers are way too
> small.
>
> The BHR is VERY similar to the v$db_cache_advice view.

I'd argue that actually the BCHR *may* be more useful than many give credit for in this sense and this sense only.

If for months your ratio hovers around the 80% mark and then it suddenly move significantly - say to 95% or to 655 - then *something* has changed and that may well be worth investigating. Note I don't say that you can tell whether the change is good or bad, or that one figure is *better* than another, just that your IO pattern has changed and it might be an idea to see why. I'm not sure that db_cache_advice would flag up the change in quite the same way. I like db_cache_advice in the sense that it shows the diminishing returns from allocating more and more ram to a cache, but it still is predicated on the assumption that your system is generating the least amount of IO that it needs to and therefore moving IO from disk to memory is a Good Thing (TM). Hands up everyone with efficient applications ....thought so.

-- 
Niall Litchfield
Oracle DBA
Audit Commission UK