Re: "Elevator ordering" and other oddities

Bryn Paul Arnold Jones (
Sat, 10 Aug 1996 17:47:29 +0100 (BST)

On Fri, 9 Aug 1996, Rakesh Dubey wrote:

> >
> > Sorry for the interruption, but there's some really excellent
> > information coming out of the I/O scheduling discussions. Problem is
> > that I don't understand a lot of it.
> >
> > Can anyone point me to a good "newless cluebie" introduction to, at
> > the least, the terminology. I've no formal CS training (high school
> > doesn't count), but I'm reasonably familiar with the kernel internals.
> >
> > Again, thanks.
> I believe that "The Design of BSD 4.3" has some details on elevator
> algorithm for ordering disk accesses.
> The real idea is simple and uses the fact that it takes considerably
> more time to position the disk head than to actually transfer the
> data. So if we can minumize disk head movements we would save
> time. Now look at a stream of I/O requests and reorder it so that the
> accesses are towards the center of the disk (i.e. decreasing cylinder
> number) and then after a point towards the periphery of the disk. In
> order, you go up and then down.

I think you just described saw tooth ...

> -Rakesh

PGP key pass phrase forgotten,   \ Overload -- core meltdown sequence 
again :( and I don't care ;)      |            initiated.
                                 / This space is intentionally left   
                                |  blank, apart from this text ;-)