Contrary to you I don't think *SIMULATING* aio the same as aio.
Also, contrary to you, I did test. It was faster without disk_async_io
= true on vxfs.
As far as I am concerned, you are simply incorrect: the article is
clearly stating the Oracle kernel *simulates* aio (after the initial
kaio call failed). I have demonstrated it actually was slower. Not sure
why you insist on contradicting me.
I usually don't make up fairy tales.
--
Sybrand Bakker
Senior Oracle DBA
Received on Thu Jun 23 2005 - 06:42:46 CDT
