Doug Ledford wrote:
>
> Mark Veteikis wrote:
> >
> > > Intel would release a Linux version of IOMETER. It's a great measurement tool.
>I've
> > > sent them several e-mails asking them to port it to Linux at [EMAIL PROTECTED] -
> > > maybe if others would to it would help. Also all disk access is raw there is no
> >
> > I specifically asked Intel about this port, and when it was going to be done.
> > This at an Intel Developers Forum about a year ago, in their "We back Linux"
> > marketing push. Being in a public forum they probably spoke a little too early
> > and stated that the Linux port was coming.
> >
> > The problem with I see with IOMETER is that if Intel ports it to Linux and
> > a naive user starts running benchmarks on identical hardware with both NT and
> > Linux, somebody's feelings might just get hurt. :)
>
> IOMETER won't be on linux for a while still yet. It relies heavily upon both
> Raw I/O and Async I/O. We don't currently support those two together.
The scsi generic driver has done async IO since lk 2.2.6 . From lk
2.3.43 (sg version 3) it is capable of doing raw I/O using Stephen
Tweedie's kiobuf internal interface. That code is "ifdef"-ed out
until various issues with kiobuf are resolved (I posted a list of
concerns to the linux kernel newsgroup some months ago). Sct said
that he is getting to them but he is quite busy with other things
at the moment I expect.
So IOMETER could be ported now. The kiobuf issue is an internal
kernel matter that could be bypassed, if necessary. I would prefer
to wait for Stephen.
Doug Gilbert
-
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to [EMAIL PROTECTED]
