Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Usenet -> c.d.o.server -> Re: Parallel Query Question

Re: Parallel Query Question

From: Billy Verreynne <vslabs_at_onwe.co.za>
Date: Wed, 23 Feb 2000 14:40:38 +0200
Message-ID: <890kg7$6bm$1@ctb-nnrp1.saix.net>


Steve Adams wrote in message <38b35673.508042256_at_news.eagles.bbs.net.au>...

>I don't think this is right. Parallel table scans work by getting each
>slave to scan a rowid range using an SQL statement with the ROWID hint.
>Unless there is some equivalent hint to do a partial fast full scan, I
>don't see how a fast full scan could be parallelized. Am I missing
>something?

Fast full index scans are done in parallel - when you get them to work of course.

Let's say you have a VLT that is index on a 7 digit code. The first 3 digits indicate a customer group or region or something. OK, you can use a full table scan to scan the entire VLT in parallel to find all customers within a specified customer groups/region/whatever.

Or you can use the index instead.
SELECT
  *
FROM foo_vlt
WHERE seven_digit_code BETWEEN 1230000 AND 1239999

This will return all rows for customer group 123. The problem is that normally the index range scan will be done in serial. Doing this on a large index is a serious problem - which is why a parallel FTS is faster. However, Oracle does address the index range scan problem by allowing you to scan the index in parallel using a FFS (fast fullindex scan).

What you thus need to do is add the FFS_INDEX hint into your SQL SELECT statement and see how fast that is, compared to use a FTS.

Billy Received on Wed Feb 23 2000 - 06:40:38 CST

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US