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: Tkprof output question

Re: Tkprof output question

From: Jonathan Lewis <jonathan_at_jlcomp.demon.co.uk>
Date: Thu, 25 Nov 2004 20:28:39 +0000 (UTC)
Message-ID: <co5f9n$7hd$1@sparta.btinternet.com>

Note in-line

-- 
Regards

Jonathan Lewis

http://www.jlcomp.demon.co.uk/faq/ind_faq.html
The Co-operative Oracle Users' FAQ

http://www.jlcomp.demon.co.uk/seminar.html
Optimising Oracle Seminar - schedule updated Sept 19th





<sybrandb_at_yahoo.com> wrote in message 
news:a1d154f4.0411250633.23ac34d6_at_posting.google.com...

>
> It doesn't use any index. Given the tiny number of rows, it is
> debatable whether using any index would be advantageous. You'll need
> at least 3 I/O requests for any index lookup,
Two at a minimum for an indexed table access - index root (= leaf) block, table block.
> and probably
> you can
> read the entire table in memory with one single multiblock read.
>
But you first need to acquire the segment header, so it could be two physical - and it WILL be a minimum of three logical I/Os as the segment header is acccessed twice before a tablescan (or index fast full scan). Your premise that it is debatable is still correct - but for frequent lookups, the debate swings heavily in favour of using an index - especially if you are doing a simple PK look up. Then even more in favour of an IOT, and then if the circumstances are right a single table hash cluster.
> --
> Sybrand Bakker
> Senior Oracle DBA
Received on Thu Nov 25 2004 - 14:28:39 CST

Original text of this message

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