Comments on: How to Create Multiple Indexes in a Single DDL Statement http://oraclesponge.wordpress.com/2008/03/26/how-to-create-multiple-indexes-in-a-single-ddl-statement/ Oracle Data Warehouse Design and Architecture Fri, 04 Jul 2008 18:47:16 +0000 http://wordpress.org/?v=MU By: David Aldridge http://oraclesponge.wordpress.com/2008/03/26/how-to-create-multiple-indexes-in-a-single-ddl-statement/#comment-50885 David Aldridge Mon, 31 Mar 2008 15:39:16 +0000 http://oraclesponge.wordpress.com/?p=360#comment-50885 I'm not sure if anyone's using that command much, Mathew. There are so many limitations on it that I don't. I feel like it's one of those features where all of the underlying technologies are there, they just have to be stitched together. I’m not sure if anyone’s using that command much, Mathew. There are so many limitations on it that I don’t.

I feel like it’s one of those features where all of the underlying technologies are there, they just have to be stitched together.

]]> By: Mathew Butler http://oraclesponge.wordpress.com/2008/03/26/how-to-create-multiple-indexes-in-a-single-ddl-statement/#comment-50884 Mathew Butler Mon, 31 Mar 2008 13:18:32 +0000 http://oraclesponge.wordpress.com/?p=360#comment-50884 Sounds like a possible enhancement request against CREATE SCHEMA: http://download.oracle.com/docs/cd/B19306_01/server.102/b14200/statements_6014.htm#i2154127 The documentation states that parallel hints are supported but will be ignored - though only documented for tables. It's not clear whether parallel creation of a single index is allowed. Most of the contents of the command would have to be executed sequentially, for dependency reasons. I can't see any reason why multiple indexes from the statement couldn't be created in parallel though. As you say, enhancement request time... Sounds like a possible enhancement request against CREATE SCHEMA:

http://download.oracle.com/docs/cd/B19306_01/server.102/b14200/statements_6014.htm#i2154127

The documentation states that parallel hints are supported but will be ignored - though only documented for tables. It’s not clear whether parallel creation of a single index is allowed. Most of the contents of the command would have to be executed sequentially, for dependency reasons.

I can’t see any reason why multiple indexes from the statement couldn’t be created in parallel though.

As you say, enhancement request time…

]]> By: David Aldridge http://oraclesponge.wordpress.com/2008/03/26/how-to-create-multiple-indexes-in-a-single-ddl-statement/#comment-50876 David Aldridge Thu, 27 Mar 2008 17:55:21 +0000 http://oraclesponge.wordpress.com/?p=360#comment-50876 I have another thought on this -- in some ways this is similar to multitable insert. You can reduce the impact of running multiple single table inserts from the same source tables by relying on caching, but ultimately multitable insert is still a valuable tool I have another thought on this — in some ways this is similar to multitable insert. You can reduce the impact of running multiple single table inserts from the same source tables by relying on caching, but ultimately multitable insert is still a valuable tool

]]> By: David Aldridge http://oraclesponge.wordpress.com/2008/03/26/how-to-create-multiple-indexes-in-a-single-ddl-statement/#comment-50875 David Aldridge Thu, 27 Mar 2008 17:18:02 +0000 http://oraclesponge.wordpress.com/?p=360#comment-50875 Yes, and I have done that before, but ... i) It's a real pain to code ii) there are still the logical reads required iii) all of those buffer busy waits don't look good. http://oraclesponge.wordpress.com/2008/03/05/buffer-busy-waits/ iv) No parallelism -- although i recall hearing that using the cache hint with an parallel query does load blocks to the buffer cache, so you could start a parallel cached full table/partition scan and follow it up with simultaneous sessions building indexes noparallel (I'm not sure if parallel query on a table set to cache does the same thing). You'd need a buffer cache that was (probably considerably) larger than the data segment you're scanning. On the other hand a cache in the hardware would help anyway. Yes, and I have done that before, but …
i) It’s a real pain to code
ii) there are still the logical reads required
iii) all of those buffer busy waits don’t look good. http://oraclesponge.wordpress.com/2008/03/05/buffer-busy-waits/
iv) No parallelism — although i recall hearing that using the cache hint with an parallel query does load blocks to the buffer cache, so you could start a parallel cached full table/partition scan and follow it up with simultaneous sessions building indexes noparallel (I’m not sure if parallel query on a table set to cache does the same thing). You’d need a buffer cache that was (probably considerably) larger than the data segment you’re scanning. On the other hand a cache in the hardware would help anyway.

]]> By: Jason Bucata http://oraclesponge.wordpress.com/2008/03/26/how-to-create-multiple-indexes-in-a-single-ddl-statement/#comment-50874 Jason Bucata Thu, 27 Mar 2008 16:36:39 +0000 http://oraclesponge.wordpress.com/?p=360#comment-50874 Wouldn't buffer caching give you most of the benefit here, assuming you ran your index creates simultaneously in different sessions without parallelism? If it's partitioned, you could build index partitions one at a time... Wouldn’t buffer caching give you most of the benefit here, assuming you ran your index creates simultaneously in different sessions without parallelism?

If it’s partitioned, you could build index partitions one at a time…

]]>