RE: db_file_multiblock_read_count causing full scans to takelonger?

  • From: Kevin Lidh <kevin.lidh@xxxxxxxxx>
  • To: "Allen, Brandon" <Brandon.Allen@xxxxxxxxxxx>
  • Date: Tue, 19 Dec 2006 14:38:00 -0700

That was just an example from the two trace files.  I wanted to show the
difference in times retrieving the exact same blocks.  The overall time
was about 3 times longer so it was very consistent.  I also ran the 16
before the 128 so it should have favored the 128.  I bounced the
database (needlessly, I'm sure) before each test.  For some reason, with
my systems on my SAN, when I get to 256k, performance degrades
dramatically and not just with Oracle.  The "time dd" tests are very
consistent.  Any block size up to 255k performs the same.  Once you get
to 256k and beyond, the time increases 3x to 4x.

On Tue, 2006-12-19 at 14:24 -0700, Allen, Brandon wrote:
> That's a very small sample though - do you see that behaviour
> consistently?
> Maybe the blocks were cached by the OS or SAN when you ran with
> dbfmbrc=16, but they actually had to be read from physical disk when you
> ran with dbfmbrc=128?


Other related posts: