Re: An oldie on Blocksize and Performance?

>BTW its the 'cache buffers lru chain' of which
>there is only one, and that applies to tablescans.
>Have you been trying to confuse us ?

|||| I'm sure it's the "cache buffers chain", but doesn't a FTS scan

    put the first I/O block on the MRU end, thereby invoking the CBC?

I am really wigged out with this thing. I moved my buffer cache up to about 1.5G, and it's using only 400MB at peak, and I wonder what damage i've done...I've set a bunch of goofy parameters ( _db_block_max_cr_cnt and the like )....so....I notice that 400MB is darn close to what i've bumped CACHE_SIZE_THRESHOLD up to, and it makes me wonder.......anyway, this is a great little test database that is going away soon, so i am playing
pretty hard with it....it was 70x too slow to begin with, and now its just 5x too
slow....not bad for no disk work, no application work, and no new db objects,
just oracle kernel tweaking....
>
>--
>
>Jonathan Lewis
>Yet another Oracle-related web site: www.jlcomp.demon.co.uk
>
>Ross Mohan wrote in message <7h5k2h$680_at_sjx-ixn1.ix.netcom.com>...
>>The tickler about v$latch_children is an excellent point, worthy of the
>>B & B. I had thought there was only one CBC latch, and so the kids don't
>>make any
>>sense to me....Could you say a bit more about the children in this case?
>>
>>Hash buckets r us, so I don't think I've gotten to the core
>>of the problem yet. Sad truth is, i don't yet have a feel for the
multiple
>>buffer pool in 8 yet, and suspect i am flushing madly in the first 64K of
>>the CBC due to wild full table scans....but, this is a guess...i need,
>>somehow
>>to "instrument" my buffer cache to nail this one down....
>>
>>Sorry, JL, but plane fare's your worry! :) I won't order your Newcastle
or
>>Smithick's
>>until you land.
>
>
>
Received on Tue May 11 1999 - 22:03:11 CDT