Re: 10G and UFS - long write times

  • From: "Alex Gorbachev" <gorbyx@xxxxxxxxx>
  • To: RROGERS@xxxxxxxxxxxxx
  • Date: Thu, 8 Jun 2006 21:32:05 +0200

Sorry, but what it has to do with 10g specifics? The difference is
that you fill your redo logs and archivers do not keep up with LGWR.
If you IO subsystem is performant enough, you can consider adding more
archivers and more redo log groups (you have more than 2-3. do you?).
To detect if you have ARCx performace problem - check V$LOG - column
NEED_ARCHIVING (or something) should be YES for all/most of redo logs
during loading.

If your LGWR gets affected by archiver - this is IO subsystem
saturation - make sure your archive log destination on different
physical disks and that you at least alternate your redo logs between
two disks.

Also consider direct load usin SQL*Loader (but don't forget
compromised recoverability).

 That is exactly what I noticed when I migrated to 10G on Linux. The
log waits increased while writing the archivelogs. I have 2 systems,
same OS same 10g setup but one does not do archiving. The archiving
system waits during a SQLLDR load and the non archivelogging does not
wait. I checked Metalink at the time and it did not explain the problem.
I have sence upgraded to 10g2 and the wait is still there.

-- Best regards, Alex Gorbachev

http://oracloid.blogspot.com
--
//www.freelists.org/webpage/oracle-l


Other related posts: