RE: Anyone using multi-block sizes for their databases

  • From: "Kevin Closson" <kevinc@xxxxxxxxxxxxx>
  • To: <juancarlosreyesp@xxxxxxxxx>, "Lex de Haan" <>
  • Date: Thu, 4 Aug 2005 11:34:02 -0700

I had a paper published in the proceedings of
Computer Measurement Group and Oracle Internals
Magazine on this topic. Well, not exactly
this topic, but multiple buffer pools to be
exact so they are orthoganal in my opinion

anyone want to see it? it is 247 bytes PDF
and I don't think I am supposed to email that to the
list so ....

>>>-----Original Message-----
>>>From: oracle-l-bounce@xxxxxxxxxxxxx 
>>>[mailto:oracle-l-bounce@xxxxxxxxxxxxx] On Behalf Of Juan 
>>>Carlos Reyes Pacheco
>>>Sent: Monday, July 25, 2005 6:37 AM
>>>To: Lex de Haan
>>>Cc: Oracle-L@xxxxxxxxxxxxx
>>>Subject: Re: Anyone using multi-block sizes for their databases
>>>Really Thanks Lex for the point is very important, I talk 
>>>with tom and I change my advice, what do you think about this.
>>>So the point is,
>>>Multiple block sizes were implemented to transport tablespaces.
>>>Don't use multiple block sizes, unless you have evidence 
>>>this really improves your performance, and this don't makes 
>>>unmanageable your database.
>>>Using multi-block tablespace block sizes, can incrase the 
>>>complexity of your cache maintenance, in a way you can have 
>>>to go back and return to a single block size tablespace.
>>>Remember neither 10g automatic sga tuning feature includes 
>>>different block sizes for autotuning.
>>>For me, I have a small database, I have a 32k tablepace for 
>>>blob documents and 16k for indexes. Because this allows to 
>>>separate the memory. I'm in process of test, I like this 
>>>idea because I can say "blob documents don use more than 10 
>>>MB of memory, and I set a fixed memory only for indexes."
>>>But I can't say this really improves my performance. I'll 
>>>have to run a more serious test to get a conclusion, 
>>>sometime in the future :).

Other related posts: