Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Usenet -> c.d.o.server -> Re: low buffer_hit_rate !!!!!

Re: low buffer_hit_rate !!!!!

From: Doug Cha <dcha_at_shell2.ba.best.com>
Date: 1997/12/30
Message-ID: <68aa9k$vb$3@nntp1.ba.best.com>#1/1

Timo Haatainen <Timo.Haatainen_at_tietogroup.com> wrote:
: > A full table scan fills up the cache , and loads the data
: > buffers. Thereby if you have a full table scan there is a
: > posibility that you should be getting more hits on
: > the buffer since the data is already there ...(you did a full
: > table scan remember) .....Provided of course that you have
: > adequate large buffers to accompany the data and they are not
: > aged out ....
 

: The data from a full table scan is loaded to the end of the LRU list of
: buffer cache and therefore is aged out first. I think there is some
: parameter with which you can tell Oracle which full table scans are
: treated as long tables and which are short tables. Anyway, this is the
: mechanism to prevent 'bad selects' to flush the whole buffer cache.

I recall there is some flexibility in controlling the cache features with the +cache sql hints. Check your SQL reference guide for this. It is definitely something I recall clearly from the O8 Server Concepts guide.

: When scanning a long table Oracle writes n bytes to the end of the LRU
: list of cache, reads n bytes more, flushes n bytes from cache to have
: space in cache (these are flushed from the end of the list), writes n
: bytes to cache, reads another n bytes, ....

What determines the 'n' number of bytes? Received on Tue Dec 30 1997 - 00:00:00 CST

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US