RE: 2GB or not 2GB (datafile limit)? That is the question.

  • From: "Kevin Closson" <kevinc@xxxxxxxxxxxxx>
  • To: <oracle-l@xxxxxxxxxxxxx>
  • Date: Fri, 8 Sep 2006 09:55:03 -0700

>>>     I don't know about OS issues, but we also stick to a max 2GB
datafile size for a different reason.  Almost all our servers (Linux and
HPUX) are on one of our SANs and we've found that the OSs tend to give
each mount point the same (on average) fraction of I/O bandwidth to the
SANs.  This was startlingly obvious when, at the recommendation of one
of our SysAdmins, we put a Production database all under 1 mount point
and its performance suffered greatly - no detailed analysis, it was just
obvious.  When we spread that same database across multiple mount points
on the same server and SAN, performance improved dramatically. 

...that is not so weird, but you didn't say what OS.  There certainly
have been and
are OS/HBAs that limit queue depth on a per-lun basis. It depends on the
OS/HBAs ... 
Personally, I think it would make more sense in that SPECIFIC case to
have a RAID 1+0 volume
of those many 2GB chunks and then put an Oracle-optimized filesystem on
it and
use the Oracle file sizes of your opposed to limiting
Oracle's file sizes
to suit the peculiarities of your SPECIFIC san setup...

SPECIFIC is caps on purpose...way too much folklore out there...


Other related posts: