io scheduler silly question perhaps..

io scheduler silly question perhaps..

[Dave Airlie]

I have an embedded system which has two read-only flash devices (one a
PIO ATA flash disk, and one MDMA capable flash)

As I'm doing no writing in this system and most of my reads are sequential
(streaming movies or images) would my choice of io scheduler be very
important?

Regards,
Dave.

-- 
David Airlie, Software Engineer
http://www.skynet.ie/~airlied / airlied at skynet.ie
Linux kernel - DRI, VAX / pam_smb / ILUG

Re: io scheduler silly question perhaps..

[Nate Diller]

Try benchmarking Anticipatory or Deadline against Noop, preferably
with your actual workload.  Noop is probably what you want, since
there is not much use in avoiding large "seeks".  It could be though
that request merging, which the non-noop schedulers all perform, willl
cause Noop to lose.  I haven't tried any I/O scheduler benchmarks with
flash, but perhaps we need a simple "merge only" scheduler for this
sort of thing.

Let me know what the results are.

NATE

On 7/28/05, Dave Airlie <airlied@linux.ie> wrote:
> 
> I have an embedded system which has two read-only flash devices (one a
> PIO ATA flash disk, and one MDMA capable flash)
> 
> As I'm doing no writing in this system and most of my reads are sequential
> (streaming movies or images) would my choice of io scheduler be very
> important?
> 
> Regards,
> Dave.
> 
> -- 
> David Airlie, Software Engineer
> http://www.skynet.ie/~airlied / airlied at skynet.ie
> Linux kernel - DRI, VAX / pam_smb / ILUG
> 
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
>

Re: io scheduler silly question perhaps..

[Jens Axboe]

On Thu, Jul 28 2005, Nate Diller wrote:
> Try benchmarking Anticipatory or Deadline against Noop, preferably
> with your actual workload.  Noop is probably what you want, since
> there is not much use in avoiding large "seeks".  It could be though
> that request merging, which the non-noop schedulers all perform, willl
> cause Noop to lose.  I haven't tried any I/O scheduler benchmarks with
> flash, but perhaps we need a simple "merge only" scheduler for this
> sort of thing.
> 
> Let me know what the results are.

deadline is the appropriate choice, you could still have read starvation
issues with noop. anticipatory doesn't make any sense, as the device has
no seek penalty.

and hey, don't top post! now we lost daves original mail.

-- 
Jens Axboe

Re: io scheduler silly question perhaps..

[Nate Diller]

On 7/29/05, Jens Axboe <axboe@suse.de> wrote:
> On Thu, Jul 28 2005, Nate Diller wrote:
> > Try benchmarking Anticipatory or Deadline against Noop, preferably
> > with your actual workload.  Noop is probably what you want, since
> > there is not much use in avoiding large "seeks".  It could be though
> > that request merging, which the non-noop schedulers all perform, willl
> > cause Noop to lose.  I haven't tried any I/O scheduler benchmarks with
> > flash, but perhaps we need a simple "merge only" scheduler for this
> > sort of thing.
> >
> > Let me know what the results are.
> 
> deadline is the appropriate choice, you could still have read starvation
> issues with noop. anticipatory doesn't make any sense, as the device has
> no seek penalty.

I'm not sure I understand your concern with noop.  even if there were
writes in the queue (he said there were none/few) i don't see how read
starvation could occur with a FIFO.  also, I have read that for large
flash devices, there is a (small) seek penalty, but that it's
relatively constant for large vs. small seeks.  can someone elaborate
on that?

so dave, try deadline vs noop and let us know what you get.  also, are
you concerned with CPU use?

> 
> and hey, don't top post! now we lost daves original mail.

yeah, you're right, so much for gmail "quick reply"

NATE
> 
> --
> Jens Axboe
> 
>