Re: ASM Record Deletions

  • From: "Alex Gorbachev" <gorbyx@xxxxxxxxx>
  • To: "Kevin Closson" <kevinc@xxxxxxxxxxxxx>
  • Date: Thu, 15 Jun 2006 21:34:05 +0200

>>>ASM assumes that your design is good (and so does RAC, for
>>>example) but Oracle doesn't mention it much.

Now this one confuses me. ASM keeps track of where blocks
are on disk. It is ignorant about content. Why do you pose
that is assumes anything ?

Well, to formulate it correctly, ASM developers assumed that your design is good enough to avoid having few single hot 1M or 128K chunks. They also assumed that your disks deliver equal performance. Having your data distributed over the disks equally and no hot chunks, Oracle should hit all the disks with about the same number of IOs. So if we have 100 GB diskgroup and 20 GB are "hot" than chances are that all these 20G are distributed equally over all disks in the diskgroup.

In the extream case, if you have 20MB diskgroup and 2 MB are hot than
chances are that these 2 MB will end up on the same disk.

If as any good DBA you know your application and IO patterns pretty
well you might actially want to put those 20 GB separately on separate
hot disks but than another DG is the solution. But this is another
story.

--
Best regards,
Alex Gorbachev

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


Other related posts: