On Sat, 2006-02-18 at 13:55 -0800, Brad Fuller wrote:
> Lee Revell wrote:
> >
> > This helps a lot:
> >
> > echo 64 > /sys/block/hdc/queue/max_sectors_kb
> >
> > (replace hdc with your drive, repeat for all of them)
> >
> > Try that before you resort to patching the kernel.
> >
> > IIRC PD is not the best designed RT app (GUI stuff can delay the RT
> > threads), but this might have improved lately.
> >
> What is given up by setting this to 64?
Presumably, a little bit of disk performance. Low latency always
involves giving up some performance, because it's a matter of doing some
kind of work in smaller chunks. I have not measured it.
> I have a sata set at 1024 and an
> IDE at 128. Certainly they were set at these settings for a reason (I
> hope a good reason)
>
Heh, that's a funny story - it's actually the farthest thing from a good
reason I can imagine. The maximum IO size is determined by whether
48-bit addressing is in use which is determined by - wait for it - the
size of the drive! The sata drive is bigger than the IDE right?
Lee
Received on Sun Feb 26 20:15:24 2006
This archive was generated by hypermail 2.1.8 : Sun Feb 26 2006 - 20:15:24 EET