Re: Index compression on Oracle 9.2

  • From: "jaromir nemec" <jaromir@xxxxxxxxxxxx>
  • To: <gogala@xxxxxxxxxxxxx>, <Christian.Antognini@xxxxxxxxxxxx>
  • Date: Sun, 5 Feb 2006 23:40:41 +0100

Hi,



I must say that the paper is excellent, ..



Indeed, as other papers from this source.



.. especially the finding that ..



I encountered some additional "minor issues" connected with compression of tables and indexes; may be some one will find this interesting:




* the AVG_ROW_LEN is the length of the *uncompressed* row, so this value doesn't change after the compression.

* a little bit tricky is the compression on the composite (e.g. range / list) partitioned table if the intention is to insert the data uncompressed and to compress them later on.



* on a table with composite partition is not possible to direct compress the partition or subpartition with an alter table statement

ALTER TABLE t MOVE (sub)partition x compress; -- gives ora-14160 / ora-14257



* the workaround is to flag the whole table as compressed and move the subpartition without compress.

alter table t compress;

alter table t move subpartition x; -- this subpartition will be compressed



* the bottom line is that on a table with composite partitioning you may control the compression on the subpartition level (e.g. current part of the table is not compressed; history is compressed) but the COMPRESSION column in dba_tab_subpartition doesn't say anything about the actual status of the compression of the subpartition.

* until 10g (not sure exact what release; it works in 10r2) it is not possible to define a local compressed index on a composite partitioned table (ORA-08113)



Regards,



Jaromir D.B. Nemec
----- Original Message ----- From: "Mladen Gogala" <gogala@xxxxxxxxxxxxx>
To: <Christian.Antognini@xxxxxxxxxxxx>
Cc: <j.velikanovs@xxxxxxxxx>; "ORACLE-L" <oracle-l@xxxxxxxxxxxxx>
Sent: Friday, February 03, 2006 2:14 PM
Subject: Re: Index compression on Oracle 9.2




On 02/03/2006 07:18:51 AM, Christian Antognini wrote:
Jurijs

>BTW: If some body have got results of that kind of testing, please do
>not hesitate to share     them with the list ;))))






-- //www.freelists.org/webpage/oracle-l


Other related posts: