RE: Choosing data file size for a multi TB database? (just the bit about 90% utilization)

  • From: "Mark W. Farnham" <mwf@xxxxxxxx>
  • To: <DGoulet@xxxxxxxx>, <Thomas.Mercadante@xxxxxxxxxxxxxxxxx>, <BranimirP@xxxxxxxx>, <oracle-l@xxxxxxxxxxxxx>
  • Date: Tue, 30 Aug 2005 10:52:37 -0400

<snip>

your OS vendor.  HP I know is not happy if you exceed 90% utilization on
a mount point.
<snip>

I'm not sure what you mean by "not happy" but the default default inode plus
bad spot allocation on many UNIX descendents is still 10%.

Most badspot mapping is done in the hardware now, and 10% has become a
ridiculous amount of storage to dedicate to information about the
name, dlm, size, etc., starting point, and various other metadata about the
list of files on a mount point.

This is especially true if a mount point is dedicated to one or a modest
number of files, as should be the case on mount points holding
Oracle data files.

So, IF the unhappiness you're referring to is this 10% reservation, then by
all means build your mountpoints for Oracle data files with much less
allocated to inode space. The exact syntax may vary by platform, so I'm not
even going to start. If you converse with your Hardware/OS vendor, be
certain they understand the question or you'll just get back "Oh, you
shouldn't monkey with that" or some other such nonsense. 10% is a whole lot
to toss down the drain for inode storage.

Then again, you might be writing about something entirely different.

Regards,

mwf


--
//www.freelists.org/webpage/oracle-l

Other related posts: