RE: sequential read on full-table scan?
- From: "Christian Antognini" <Christian.Antognini@xxxxxxxxxxxx>
- To: <sjaffarhussain@xxxxxxxxx>
- Date: Mon, 23 May 2005 12:02:33 +0200
>>Chained or migrated rows It is a problem if you see many db file
>>sequential read waits against a table when the execution plan of the
>>SQL statement calls for a full table scan. This indicates the table
>>has many chained or migrated rows. Oracle goes after each chained or
>>migrated row with the single-block I/O call.
>
>IMO this is not true.
A clarification... In case of migrated rows the additional I/O isn't =
performed because the row header is skipped. The addition I/O, i.e. the =
WAIT event, is present in the tracefile if a row is chained and 1) it is =
not chained in the block itself (more than 255 columns) 2) the chained =
block has already been read and it's still in the buffer cache. If a row =
is migrated and chained at the same time, the "rules" for chained rows =
applies.
Chris
--
//www.freelists.org/webpage/oracle-l
Other related posts:
