| Oracle FAQ | Your Portal to the Oracle Knowledge Grid | |
 |
 | |||
| Oracle FAQ | Your Portal to the Oracle Knowledge Grid | |
|
| |||
Home -> Community -> Mailing Lists -> Oracle-L -> RE: recycle and keep buffer in other tablespace block sizes
Hi Mark, thanks,
I was thinking too to in run a test, this year, to move all indexes to a 16k
tablespace, to see if there is some improvement.
But why don't do it as a group in this list.
If we do test as a group I t hink we could find very very interesting things
becaues the big problem is never is enough time to test everything.
The question will be interesting to set in this forum a goal and those who are interesting joing to it.
For example something like
subject: All who wants to test 16k tablespaces
The limit to send you trace files is (one week from today)
The goal of this test is to create a tablespace using 16, and compare the
results in
small tables 1-100k
medium tables 100k 10;M
Big tables more than 10M
You have to post the results from a trace file in xxx debug level. Then who started this goal, copies all this to a paper, some comments a post in the forum a link to that test.
I don't know if you like this idea, what do you think.
-------Original Message-------
From: oracle-l_at_freelists.org
Date: 07/21/04 11:37:00
To: oracle-l_at_freelists.org
Subject: RE: recycle and keep buffer in other tablespace block sizes
Since Oracle has emphatically and repeatedly stated they do not intend varying block sizes as a performance tool, I believe they intentionally did not create keep and recycle pools for alternate sizes in a database. And they are moving toward overall target sizes rather than user specified allocations.
Although I have some theories I'm researching regarding using alternate block sizes for performance purposes in 10g, I'll refrain from reporting results until I have a PDB (pretty damned bulletproof) example set. If the results prove simple enough to manage and of substantial benefit, adding those allocations might seem reasonable (although it pretty much flies in the face of the trend toward using a target size and having the actual usage of the instance drive dynamic self tuning.)
I haven't seen a thread on experimental results of the effects of using multiple block sizes on throughput. If anyone has such results to share, I wouldn't mind skipping the reinvention of the wheel. (Please note I'm not asking for new speculations on likely to work cool tricks to benefit performance with multiple block sizes -- I've already got a longer list than I really want to test.) Also, I'm more interested in thoughtful attempts to establish that benefit can be gained (whether successful or not) than in the infinite number of cases that can easily be concocted to show no difference.
(The stated intent of supporting multiple block sizes is to support transportable tablespaces amongst databases with varying block sizes.)
mwf
-----Original Message-----
From: oracle-l-bounce_at_freelists.org
[mailto:oracle-l-bounce_at_freelists.org]On Behalf Of Juan Carlos Reyes
Pacheco
Sent: Wednesday, July 21, 2004 9:58 AM
To: oracle-l_at_freelists.org
Subject: recycle and keep buffer in other tablespace block sizes
Hi list I'm investigating this,
if your default blocksize is 8k, and you create a tables using 2k block
you have the
ALTER SYSTEM SET DB_2K_CACHE_SIZE = 100M;
to set the size for this block cache
But how do you set the keep and recycle size for this block size
There are none of this parameter.
ALTER SYSTEM SET DB_2K_KEEP_CACHE_SIZE = 1M;
ALTER SYSTEM SET DB_2K_RECYCLE_CACHE_SIZE = 1M;
Thank you
Juan Carlos Reyes Pacheco
OCP
FAQ is at http://www.freelists.org/help/fom-serve/cache/1.htmlput 'unsubscribe' in the subject line.
-----------------------------------------------------------------
----------------------------------------------------------------
Please see the official ORACLE-L FAQ: http://www.orafaq.com
----------------------------------------------------------------
To unsubscribe send email to: oracle-l-request_at_freelists.org
 |
 |