Re: disk head scheduling
Arvind Sankar (arvinds@mit.edu)
Thu, 18 Mar 1999 16:05:26 -0500
On Thu, Mar 18, 1999 at 08:50:36PM +0100, Ingo Molnar wrote:
>
> On Thu, 18 Mar 1999, Arvind Sankar wrote:
>
> > I have a currently working algo which is greedy instead. i.e. for a new
> > request, it computes the place where its insertion will produce the least
> > incremental cost (where cost is the number of additional sectors to seek).
> > It seems to work for me. No idea how to compare the two algos, though. Can
> > somebody help me out here...
>
> be careful. These are all well-known problems, and your algorithm (also
> called SSF (smallest-seek-first) scheduling) is not fair enough. It has to
> be coupled with a mechanizm that provides fairness. (ie. guarantees that a
> request will not 'hang around' for too long time, unfairly blocking a
> process just because the position of the sector is unfortunate)
yeah. besides, it isn't really better than elevator, since doing reads while
going backward on a disk is worse than just going to the start and doing all
of them in the forward direction.
Reverting now :)
>
> > Another point is that IN_ORDER seems to be called only for two requests
> > on the same device, so no idea why it compares the device numbers.
>
> no, all requests (for all devices) are in a single 'queue'. (Per-major
> device queues is candidate 2.3 feature, it's really simple)
oh. I realized that it was at least per-major, but it's global, huh?
-- arvind
-
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/