Hello Roch,
Monday, July 17, 2006, 6:09:54 PM, you wrote:
R> Robert Milkowski writes:
>> Hello zfs-discuss,
>>
>> What would you rather propose for ZFS+ORACLE - zvols or just files
>> from the performance standpoint?
>>
>>
>> --
>> Best regards,
>> Robert mailto:rmilkowski at task.gda.pl
>> http://milek.blogspot.com
>>
>> _______________________________________________
>> zfs-discuss mailing list
>> zfs-discuss at opensolaris.org
>> http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
R> Not sure to what extent this would suffer from this:
R> Synopsis: large writes to zvol synchs too much, better cut down a
little
R> Number: 6428639
R> At first glance, I would think it''s not in the picture as
R> oracle won''t issue those jumbo sized writes.
I also saw some performance problem using zvols with iSCSI reported by
others (Ben?). I decided to go with files right now.
R> Right now, I''ve seen more focus of getting ZFS/Oracle
R> working well with files because that''s what we see as the
R> demand. But given that Oracle is ready to deal with raw and that
R> files bring a extra set of constraints I would think that
R> zvol is best. Plus you seem to have an early adopter frame
R> of mind ;-)
Yeah, sometimes when potential benefits outweigh risk I like to go
with ''edge'' technologies - and hey, it''s A LOT of fun
(and sleepless
nights).
For now I created raid-10 group from 14 SCSI disks in one pool, with
two datasets delegated to two zones. In each zone one Oracle instance.
Separate sub-datasets for oracle files with record size set to 32k and
8k (different Oracle files will use different db block size).
I will create another pool with FC storage - have to decide yet to do
hw raid or software raid in ZFS. Right now one of the missing things
in stable Solaris (s10u2) is ZFS Hot Spare support.
ps. our db admins already like snapshots, clones, etc :)
--
Best regards,
Robert mailto:rmilkowski at task.gda.pl
http://milek.blogspot.com