Re: ORBS Elevator...

From: Andrea Arcangeli (andrea@suse.de)
Date: Sat Jul 29 2000 - 09:39:45 EST


On Fri, 28 Jul 2000, Andre Hedrick wrote:

>Andrea,
>
>As much as I do not like the elevator, it will be a saving grace for ORBS.

You should like it. When I started playing with it I thought it wasn't
going to make an huge difference to have the elevator or not, but now
after some more time I seen it really makes an huge difference in numbers ;).

I'm sorry but I'm not sure to know what is ORBS (I also did a fast search
but I couldn't find it out). (I only know the anti spam one and I'm sure
you're not talking about it 8).

>I will get the LBA position and track pattern and do to the linear C
>nature of the media WRT where LBA 0 is defined. I will know the exact
>track/sector position on the media, thus with that discrete knowledge and
>knowing how the media is physically accessed, we can exactly model an
>orbs_elevator.
>
>Are you game to play math-model and prediction?

I guess ORBS means that you will provide the information about the
geometry of the disk so that we can elevate the request in function of the
geometry and not only in function of the geometry-blind block sector
(we're not doing that right now because the geometry is always faked).

It's not obvious to me that the additional information will help
performance significantly, but I'll be glad to give it a try and to run
some benchmark to see if it helps significantly.

Andrea

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/



This archive was generated by hypermail 2b29 : Mon Jul 31 2000 - 21:00:30 EST