Re: Direct IO

Hi Yong.

That's an excellent explanation and essentially what I was looking for. I'm no expert on the NTFS file system (I most decidedly have better things to do, as do we all), but I was unsure of the mechanics of the Direct I/O implementation on NT. What you have summarized from Steve sounds about right. I'll do some investigation and see what I can come up with. I appreciate your answer.

Cheers,

Dave Haas

"Yong Huang" <yong321_at_yahoo.com> wrote in message news:b3cb12d6.0106252027.235087b4_at_posting.google.com...
> Steve Adams has done some research on this: "Oracle on NT always
> explicitly does direct I/O. The datafiles are created with
> FILE_FLAG_NO_BUFFERING and opened with IRP_NOCACHE." It's easy to
> mistakenly extrapolate that to mean that NT never does buffering. Note
> that he only says Oracle on NT, not NT in general. And he seems to say
> even the Microsoft people think the NT porters of Oracle are doing a
> bad job in choosing direct I/O on NT since direct I/O on NT disables
> fast I/O. So don't think direct I/O is always faster.
>
> There's an O'Reily book on NT File Systems Internals, but I spent at
> most 5 hours reading it. If you think you know OS well enough, you may
> take a look and educate us here.
>
> Yong Huang
> yong321_at_yahoo.com
>
> "Dave Haas" <davidh_at_no.spam.hotmail.com> wrote in message
 news:<jdPZ6.80657$Jg4.8474809_at_news1.telusplanet.net>...
> >
> > In terms of the NT Direct IO thing. If it is on by default, does that
 then
> > mean that all operations on an NTFS device bypass the cache? That seems
 a
> > little wierd as NT likes to tout itself as a fileserver OS (with a fair
> > amount of physical memory being dedicated to file ops IIRC).
> >
> > Cheers,
> >
> > Dave
Received on Tue Jun 26 2001 - 00:31:53 CDT