>Subject: Performance effect of number of columns on a select clause?
>From: "Pasi Kovanen" Pasi.Kovanen_at_removethis.iki.fi.invalid
>Date: 2/25/02 4:35 AM Eastern Standard Time
>Message-id: <a5d0ev$s9p$1_at_news1.song.fi>
>
>We did some tests using different SELECT clauses and were
>surprised to found how much the number of columns that
>were selected affect the performance.
>
>For example, execution time of
> SELECT column1 from foobar
>seems to be almost constant whatever the size of data returned in
>column1 is 20 - 400 bytes (only 5% decrease w/ 400 bytes).
>
>Wheres
>SELECT column1, column2, ... column13 from foobar can take
>even 40% longer though most of the columns are null and
>returned amount of data is less than the 400 bytes column1 alone.
>
>Is there a way to overcome (or kludge ;) this?
>
>
>
>
>
>
>
>
Check to see that the first query did not just fetch the data from an index
rather than the table. If you plan your indexes well, that can lead to some
significant performance gains. Use EXPLAIN PLAN to see the data path.
Received on Tue Mar 05 2002 - 13:47:38 CST
